{ "export_catalog_version": "2.2.3", "export_schema_version": "1.0", "metadata": { "description": "Compact, source-grounded catalog of Apple Core AI models, artifacts, capabilities, modalities, runtime metadata, device support and sources.", "last_verified": "2026-07-04", "name": "Core AI Catalog", "repo": "coreai-catalog", "upstream": [ "coreai-model-zoo", "apple/coreai-models" ], "version": "2.2.3" }, "models": [ { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-5-0-8b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen3-5-0-8b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Qwen3.5-0.8B", "notes": "Top-1 exact vs Hugging Face reference according to upstream.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "969MB", "parameters": "0.8B", "precision": "int8", "quantization": "int8lin" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/qwen3.5.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3.5-0.8B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-5-2b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen3-5-2b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Qwen3.5-2B", "notes": "Top-1 exact vs Hugging Face reference according to upstream.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "2B", "precision": "int8", "quantization": "int8lin" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/qwen3.5.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3.5-2B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-6-35b-a3b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "moe" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Qwen", "id": "qwen3-6-35b-a3b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Qwen3.6-35B-A3B", "notes": "Mac-only MoE; custom gather_qmm Metal kernel.", "runtime": { "aot_required": false, "custom_kernel": true, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "35GB", "parameters": "35B / ~3B active", "precision": "int8", "quantization": "sym8-gather" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/qwen3.6.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3.6-35B-A3B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-6-27b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Qwen", "id": "qwen3-6-27b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Qwen3.6-27B", "notes": "Dense Mac-only LLM.", "runtime": { "aot_required": false, "custom_kernel": true, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "28GB", "parameters": "27B", "precision": "int8", "quantization": "int8hu" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/qwen3.6-27b.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3.6-27B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "glm-4-7-flash", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "moe", "mla" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "GLM", "id": "glm-4-7-flash", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "GLM-4.7-Flash", "notes": "MoE + MLA; custom gather_qmm Metal kernel.", "runtime": { "aot_required": false, "custom_kernel": true, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "30GB", "parameters": "30B / ~3B active", "precision": "int8", "quantization": "sym8-gather" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/glm-4.7-flash.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "zai-org/GLM-4.7-Flash" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "gemma-4-e2b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Gemma", "id": "gemma-4-e2b", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Gemma Terms" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Gemma 4 E2B", "notes": "Text decoder; upstream reports 8/8 exact vs HF.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "1.3GB", "parameters": "E2B", "precision": "int4", "quantization": "int4lin" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/gemma4-e2b.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "gemma-4-e4b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Gemma", "id": "gemma-4-e4b", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Gemma Terms" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Gemma 4 E4B", "notes": "Official-QAT int4.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "3.7GB", "parameters": "E4B", "precision": "int4", "quantization": "int4lin-QAT" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/gemma4-e4b.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "google/gemma-4-E4B-it-qat-q4_0-unquantized" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "gemma-4-12b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Gemma", "id": "gemma-4-12b", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Gemma Terms" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Gemma 4 12B", "notes": "Custom flash-decode kernel.", "runtime": { "aot_required": false, "custom_kernel": true, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "8.2GB", "parameters": "12B", "precision": "int4", "quantization": "int4linsym" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/gemma4-12b.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo", "apple-coreai-models" ], "status": "confirmed", "streaming": true, "upstream_repo": "google/gemma-4-12B-it-qat-q4_0-unquantized" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "gemma-4-31b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Gemma", "id": "gemma-4-31b", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Gemma Terms" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Gemma 4 31B", "notes": "Custom flash-decode kernel.", "runtime": { "aot_required": false, "custom_kernel": true, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "19GB", "parameters": "31B", "precision": "int4", "quantization": "int4linsym" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/gemma4-31b.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo", "apple-coreai-models" ], "status": "confirmed", "streaming": true, "upstream_repo": "google/gemma-4-31B-it-qat-q4_0-unquantized" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "lfm2-5-1-2b-instruct", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "instruction-following" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "LFM", "id": "lfm2-5-1-2b-instruct", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "LFM Open License v1.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "LFM2.5-1.2B-Instruct", "notes": "Hybrid architecture.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "1.6GB", "parameters": "1.2B", "precision": "int8", "quantization": "int8hu" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/lfm2.5.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "LiquidAI/LFM2.5-1.2B-Instruct" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "lfm2-5-8b-a1b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "moe" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "LFM", "id": "lfm2-5-8b-a1b", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "LFM Open License v1.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "LFM2.5-8B-A1B", "notes": "MoE path with gather_qmm.", "runtime": { "aot_required": false, "custom_kernel": true, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "8.8GB", "parameters": "8B / ~1B active", "precision": "int8", "quantization": "sym8-gather" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/lfm2.5-8b-a1b-moe.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "LiquidAI/LFM2.5-8B-A1B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "granite-4-0-h-1b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "hybrid-llm" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Granite", "id": "granite-4-0-h-1b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Granite 4.0-H 1B", "notes": "Mamba2 + attention hybrid.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "1.8GB", "parameters": "1B", "precision": "int8", "quantization": "int8hu" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/granite-4.0-h.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "ibm-granite/granite-4.0-h-1b" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "granite-4-0-h-350m", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "hybrid-llm" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Granite", "id": "granite-4-0-h-350m", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Granite 4.0-H 350M", "notes": "Same upstream card as Granite 4.0-H 1B.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "350M", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/granite-4.0-h.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "ibm-granite/granite-4.0-h-350m" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "nanbeige4-1-3b", "bundle_kind": "llm", "capabilities": [ "chat", "reasoning", "agentic", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Nanbeige", "id": "nanbeige4-1-3b", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Nanbeige4.1-3B", "notes": "Dense reasoning/agentic.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "4.3GB", "parameters": "3.93B", "precision": "int8", "quantization": "int8hu" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/nanbeige4.1-3b.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "Nanbeige/Nanbeige4.1-3B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-vl-2b", "bundle_kind": "vlm", "capabilities": [ "vision-language" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen3-vl-2b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "text" ] }, "name": "Qwen3-VL 2B", "notes": "Vision-language model.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "2.3GB", "parameters": "2B", "precision": "int8", "quantization": "int8hu" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/qwen3-vl.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3-VL-2B-Instruct" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-vl-4b", "bundle_kind": "vlm", "capabilities": [ "vision-language" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen3-vl-4b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "text" ] }, "name": "Qwen3-VL 4B", "notes": "Vision-language model.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "4B", "precision": "int8", "quantization": "int8hu" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/qwen3-vl.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3-VL-4B-Instruct" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-vl-8b", "bundle_kind": "vlm", "capabilities": [ "vision-language" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen3-vl-8b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "text" ] }, "name": "Qwen3-VL 8B", "notes": "Vision-language model.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "8B", "precision": "int8", "quantization": "int8hu" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/qwen3-vl.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3-VL-8B-Instruct" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "minicpm-v-4-6", "bundle_kind": "vlm", "capabilities": [ "vision-language" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "MiniCPM", "id": "minicpm-v-4-6", "io_contract": { "entrypoint": { "framework": "CoreAILanguageModels", "init_pattern": "let engine = try await CoreAIRunner(contentsOf: bundleURL).makeInferenceEngine()", "type": "CoreAIRunner" }, "files": { "tokenizer_ref": "gpu-pipelined/minicpmv46_vlm_decode_int8hu/tokenizer" }, "inputs": [ { "modality": "image", "name": "image", "swift_type": "URL", "tensor": { "dtype": "fp16", "layout": "NCHW", "shape": [ 1, 3, 448, 448 ] } }, { "constraints": { "max_context": 4096 }, "modality": "text", "name": "prompt", "swift_type": "String" } ], "outputs": [ { "decoding": { "detokenization": "embedded tokenizer (openbmb/MiniCPM-V-4.6, vocab 248094)" }, "name": "response", "swift_type": "String" } ], "session": { "stateful": true, "streaming": true } }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "text" ] }, "name": "MiniCPM-V 4.6", "notes": "Tiny VLM.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "sub-2B", "precision": "int8", "quantization": "int8hu" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/minicpm-v-4.6.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "openbmb/MiniCPM-V-4.6" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "gemma-4-e2b-vision", "bundle_kind": "vlm", "capabilities": [ "vision-language" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Gemma", "id": "gemma-4-e2b-vision", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Gemma Terms" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "text" ] }, "name": "Gemma 4 E2B Vision", "notes": "Image+text to text.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": true, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "E2B", "precision": "int4", "quantization": "int4linsym" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/gemma4-vl.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "unlimited-ocr", "bundle_kind": "ocr", "capabilities": [ "document-ocr" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Unlimited-OCR", "id": "unlimited-ocr", "io_contract": { "entrypoint": { "framework": "CoreAI", "init_pattern": "let vision = try await AIModel(contentsOf: visionURL); let decoder = try await AIModel(contentsOf: decoderURL)", "type": "AIModel" }, "files": { "processor_ref": "assets/recipe.json", "tokenizer_ref": "tokenizer" }, "inputs": [ { "constraints": { "max_image_px": 640 }, "modality": "image", "name": "document_image", "preprocessing": { "color_format": "RGB", "normalization": "mean=0.5 std=0.5 (all channels)", "resize": "pad to 640x640 (ImageOps.pad; pad color = mean*255)" }, "swift_type": "CGImage" }, { "modality": "text", "name": "prompt_input_ids", "swift_type": "[Int32]", "tensor": { "dtype": "int32", "shape": [ 115 ] } } ], "outputs": [ { "decoding": { "detokenization": "greedy decode to eos id 1; detokenize with tokenizer/ (LlamaTokenizerFast)", "format_selector": "fixed prompt prefix (assets/prompt_input_ids.i32) — output interleaves markdown, HTML tables and LaTeX formulas; not prompt-selectable" }, "name": "text", "swift_type": "String" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "document_image", "image" ], "output": [ "markdown", "html", "latex" ] }, "name": "Unlimited-OCR", "notes": "Document OCR to markdown; tables to HTML; formulas to LaTeX.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "stock-runner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": "unknown" }, "size": { "artifact_size": "not_published", "parameters": "not_published", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/unlimited-ocr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen2-5-omni-3b-audio", "bundle_kind": "audio-lm", "capabilities": [ "audio-understanding" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen2-5-omni-3b-audio", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "audio" ], "output": [ "text" ] }, "name": "Qwen2.5-Omni-3B Audio", "notes": "Audio understanding; describes sounds, not literal transcript.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": true, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": false }, "size": { "artifact_size": "5.1GB", "parameters": "3B", "precision": "int8", "quantization": "int8lin" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/knowledge/qwen2.5-omni-audio-understanding.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-knowledge" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen2.5-Omni-3B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "kokoro-82m", "bundle_kind": "tts", "capabilities": [ "text-to-speech" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Kokoro", "id": "kokoro-82m", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "audio" ] }, "name": "Kokoro-82M", "notes": "StyleTTS2 + iSTFTNet; 28 English voices.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "335MB", "parameters": "82M", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/kokoro-82m.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo", "coreai-model-zoo-knowledge" ], "status": "confirmed", "streaming": true, "upstream_repo": "hexgrad/Kokoro-82M" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "voxcpm-0-5b", "bundle_kind": "tts", "capabilities": [ "text-to-speech" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "VoxCPM", "id": "voxcpm-0-5b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "audio" ] }, "name": "VoxCPM-0.5B", "notes": "Diffusion TTS.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "1.0GB", "parameters": "0.5B", "precision": "int8", "quantization": "int8-LM" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/knowledge/voxcpm-tts.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-knowledge" ], "status": "confirmed", "streaming": true }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "embeddinggemma-300m", "bundle_kind": "embedding", "capabilities": [ "embedding" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Gemma", "id": "embeddinggemma-300m", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Gemma Terms" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "vector" ] }, "name": "EmbeddingGemma 300M", "notes": "Text embeddings for on-device RAG / semantic search.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "300M", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo", "sources": [ "coreai-model-zoo-readme" ], "status": "confirmed", "streaming": false, "upstream_repo": "google/embeddinggemma-300m" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-embedding-0-6b", "bundle_kind": "embedding", "capabilities": [ "embedding" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen3-embedding-0-6b", "io_contract": { "entrypoint": { "framework": "CoreAI", "init_pattern": "let model = try await AIModel(contentsOf: aimodelURL)", "type": "AIModel" }, "files": { "tokenizer_ref": "tokenizer" }, "inputs": [ { "constraints": { "max_context": 512 }, "modality": "text", "name": "text" } ], "outputs": [ { "decoding": { "embedding_dim": 1024, "format_selector": "last-token pooling, L2-normalized; MRL-truncatable to 32..1024 dims" }, "name": "embedding", "swift_type": "[Float]" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "vector" ] }, "name": "Qwen3-Embedding 0.6B", "notes": "Multilingual embeddings; last-token pooling + MRL.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "1.1GB", "parameters": "0.6B", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/qwen3-embedding.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false, "upstream_repo": "Qwen/Qwen3-Embedding-0.6B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-reranker-0-6b", "bundle_kind": "reranker", "capabilities": [ "reranking" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen3-reranker-0-6b", "io_contract": { "entrypoint": { "framework": "CoreAI", "init_pattern": "let model = try await AIModel(contentsOf: aimodelURL)", "type": "AIModel" }, "files": { "tokenizer_ref": "tokenizer" }, "inputs": [ { "constraints": { "max_context": 512 }, "modality": "text", "name": "query" }, { "modality": "text", "name": "documents" } ], "outputs": [ { "decoding": { "score_range": "0..1 — probs[1] = P(yes) = relevance" }, "name": "score", "swift_type": "Float" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "query", "documents" ], "output": [ "score" ] }, "name": "Qwen3-Reranker 0.6B", "notes": "Cross-encoder reranker.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "1.1GB", "parameters": "0.6B", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/qwen3-reranker.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false, "upstream_repo": "Qwen/Qwen3-Reranker-0.6B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-nano", "bundle_kind": "object-detection", "capabilities": [ "object-detection" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR", "id": "rf-detr-nano", "io_contract": { "entrypoint": { "framework": "CoreAIObjectDetector", "init_pattern": "let detector = try await ObjectDetector(resourcesAt: \"/rfdetr-nano_float32.aimodel\")", "type": "ObjectDetector" }, "inputs": [ { "modality": "image", "name": "image", "preprocessing": { "color_format": "RGB", "normalization": "RGB scaled to [0,1] (DetectionParameters normalizationMeans/Stds)" }, "swift_type": "CGImage" } ], "outputs": [ { "decoding": { "coordinate_convention": "boundingBox: pixel CGRect, top-left origin (decoded from 300 normalized cxcywh queries; no NMS)", "label_vocab_ref": "COCO 91-id logit columns; ObjectDetectionLabels.coco default mapping" }, "name": "detections", "swift_type": "[DetectedObject]" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "boxes", "classes", "scores" ] }, "name": "RF-DETR Nano", "notes": "Object detection; no NMS.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "108MB", "parameters": "nano", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-small", "bundle_kind": "object-detection", "capabilities": [ "object-detection" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR", "id": "rf-detr-small", "io_contract": { "entrypoint": { "framework": "CoreAIObjectDetector", "init_pattern": "let detector = try await ObjectDetector(resourcesAt: \"/rfdetr-small_float32.aimodel\")", "type": "ObjectDetector" }, "inputs": [ { "modality": "image", "name": "image", "preprocessing": { "color_format": "RGB", "normalization": "RGB scaled to [0,1] (DetectionParameters normalizationMeans/Stds)" }, "swift_type": "CGImage" } ], "outputs": [ { "decoding": { "coordinate_convention": "boundingBox: pixel CGRect, top-left origin (decoded from 300 normalized cxcywh queries; no NMS)", "label_vocab_ref": "COCO 91-id logit columns; ObjectDetectionLabels.coco default mapping" }, "name": "detections", "swift_type": "[DetectedObject]" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "boxes", "classes", "scores" ] }, "name": "RF-DETR Small", "notes": "Object detection; no NMS.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "115MB", "parameters": "small", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-medium", "bundle_kind": "object-detection", "capabilities": [ "object-detection" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR", "id": "rf-detr-medium", "io_contract": { "entrypoint": { "framework": "CoreAIObjectDetector", "init_pattern": "let detector = try await ObjectDetector(resourcesAt: \"/rfdetr-medium_float32.aimodel\")", "type": "ObjectDetector" }, "inputs": [ { "modality": "image", "name": "image", "preprocessing": { "color_format": "RGB", "normalization": "RGB scaled to [0,1] (DetectionParameters normalizationMeans/Stds)" }, "swift_type": "CGImage" } ], "outputs": [ { "decoding": { "coordinate_convention": "boundingBox: pixel CGRect, top-left origin (decoded from 300 normalized cxcywh queries; no NMS)", "label_vocab_ref": "COCO 91-id logit columns; ObjectDetectionLabels.coco default mapping" }, "name": "detections", "swift_type": "[DetectedObject]" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "boxes", "classes", "scores" ] }, "name": "RF-DETR Medium", "notes": "Object detection; no NMS.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "121MB", "parameters": "medium", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-large", "bundle_kind": "object-detection", "capabilities": [ "object-detection" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR", "id": "rf-detr-large", "io_contract": { "entrypoint": { "framework": "CoreAIObjectDetector", "init_pattern": "let detector = try await ObjectDetector(resourcesAt: \"/rfdetr-large_float32.aimodel\")", "type": "ObjectDetector" }, "inputs": [ { "modality": "image", "name": "image", "preprocessing": { "color_format": "RGB", "normalization": "RGB scaled to [0,1] (DetectionParameters normalizationMeans/Stds)" }, "swift_type": "CGImage" } ], "outputs": [ { "decoding": { "coordinate_convention": "boundingBox: pixel CGRect, top-left origin (decoded from 300 normalized cxcywh queries; no NMS)", "label_vocab_ref": "COCO 91-id logit columns; ObjectDetectionLabels.coco default mapping" }, "name": "detections", "swift_type": "[DetectedObject]" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "boxes", "classes", "scores" ] }, "name": "RF-DETR Large", "notes": "Object detection; no NMS.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "122MB", "parameters": "large", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-seg-nano", "bundle_kind": "segmentation", "capabilities": [ "instance-segmentation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR-Seg", "id": "rf-detr-seg-nano", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "masks", "boxes", "classes", "scores" ] }, "name": "RF-DETR-Seg Nano", "notes": "Instance segmentation.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "108MB", "parameters": "nano", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-seg-small", "bundle_kind": "segmentation", "capabilities": [ "instance-segmentation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR-Seg", "id": "rf-detr-seg-small", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "masks", "boxes", "classes", "scores" ] }, "name": "RF-DETR-Seg Small", "notes": "Instance segmentation.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "115MB", "parameters": "small", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-seg-medium", "bundle_kind": "segmentation", "capabilities": [ "instance-segmentation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR-Seg", "id": "rf-detr-seg-medium", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "masks", "boxes", "classes", "scores" ] }, "name": "RF-DETR-Seg Medium", "notes": "Instance segmentation.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "121MB", "parameters": "medium", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-seg-large", "bundle_kind": "segmentation", "capabilities": [ "instance-segmentation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR-Seg", "id": "rf-detr-seg-large", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "masks", "boxes", "classes", "scores" ] }, "name": "RF-DETR-Seg Large", "notes": "Instance segmentation.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "122MB", "parameters": "large", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-seg-xlarge", "bundle_kind": "segmentation", "capabilities": [ "instance-segmentation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR-Seg", "id": "rf-detr-seg-xlarge", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "masks", "boxes", "classes", "scores" ] }, "name": "RF-DETR-Seg Xlarge", "notes": "Instance segmentation.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "not_published", "parameters": "xlarge", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rf-detr-seg-2xlarge", "bundle_kind": "segmentation", "capabilities": [ "instance-segmentation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "RF-DETR-Seg", "id": "rf-detr-seg-2xlarge", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "masks", "boxes", "classes", "scores" ] }, "name": "RF-DETR-Seg 2Xlarge", "notes": "Instance segmentation.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "not_published", "parameters": "2xlarge", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/rf-detr.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "adcsr-x4", "bundle_kind": "super-resolution", "capabilities": [ "super-resolution" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "AdcSR", "id": "adcsr-x4", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Apache-2.0 + OpenRAIL++" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "image" ] }, "name": "AdcSR ×4", "notes": "Super-resolution; one-step diffusion-GAN.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "1.7GB", "parameters": "×4", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/knowledge/adcsr-super-resolution.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-knowledge" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "depth-anything-3-small", "bundle_kind": "depth", "capabilities": [ "monocular-depth" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Depth Anything", "id": "depth-anything-3-small", "io_contract": { "entrypoint": { "framework": "CoreAI", "init_pattern": "let model = try await AIModel(contentsOf: aimodelURL)", "type": "AIModel" }, "inputs": [ { "modality": "image", "name": "image", "preprocessing": { "color_format": "RGB", "normalization": "RGB scaled to [0,1]" } } ], "outputs": [ { "decoding": { "format_selector": "relative depth map" }, "name": "depth" }, { "decoding": { "format_selector": "confidence map" }, "name": "confidence" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "depth_map" ] }, "name": "Depth Anything 3 Small", "notes": "Monocular depth; small variant.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "54.5MB", "parameters": "small", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/depth-anything-3.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "depth-anything-3-base", "bundle_kind": "depth", "capabilities": [ "monocular-depth" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Depth Anything", "id": "depth-anything-3-base", "io_contract": { "entrypoint": { "framework": "CoreAI", "init_pattern": "let model = try await AIModel(contentsOf: aimodelURL)", "type": "AIModel" }, "inputs": [ { "modality": "image", "name": "image", "preprocessing": { "color_format": "RGB", "normalization": "RGB scaled to [0,1]" } } ], "outputs": [ { "decoding": { "format_selector": "relative depth map" }, "name": "depth" }, { "decoding": { "format_selector": "confidence map" }, "name": "confidence" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "depth_map" ] }, "name": "Depth Anything 3 Base", "notes": "Monocular depth; base variant.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "202MB", "parameters": "base", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/depth-anything-3.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-gpt-oss-20b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "moe" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "gpt-oss", "id": "official-gpt-oss-20b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "gpt-oss-20B", "notes": "Official-recipe conversion; stock runtime.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "stock-runner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "13GB", "parameters": "20B / ~13GB", "precision": "MXFP4", "quantization": "MXFP4" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": true, "upstream_repo": "openai/gpt-oss-20b" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-qwen3-0-6b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "official-qwen3-0-6b", "io_contract": { "entrypoint": { "framework": "CoreAILanguageModels", "init_pattern": "let model = try await CoreAILanguageModel(resourcesAt: bundleURL)", "type": "CoreAILanguageModel" }, "files": { "tokenizer_ref": "macos/tokenizer" }, "inputs": [ { "constraints": { "max_context": 4096 }, "modality": "text", "name": "prompt", "swift_type": "String" } ], "outputs": [ { "decoding": { "detokenization": "embedded tokenizer (Qwen/Qwen3-0.6B, vocab 151936); LanguageModelSession.respond(to:) returns detokenized text" }, "name": "response", "swift_type": "String" } ], "session": { "stateful": true, "streaming": true } }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Qwen3 0.6B", "notes": "Official-recipe conversion; stock runtime.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "stock-runner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "0.6B", "precision": "int4", "quantization": "palettized" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3-0.6B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-qwen3-4b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "official-qwen3-4b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Qwen3 4B", "notes": "Official-recipe conversion; stock runtime.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "stock-runner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "4B", "precision": "int4", "quantization": "palettized" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3-4B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-qwen3-8b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Qwen", "id": "official-qwen3-8b", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Qwen3 8B", "notes": "Official-recipe conversion; stock runtime.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "stock-runner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "8B", "precision": "int4", "quantization": "palettized" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3-8B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-gemma-3-4b-it", "bundle_kind": "llm", "capabilities": [ "chat", "instruction-following", "text-generation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Gemma", "id": "official-gemma-3-4b-it", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Gemma Terms" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Gemma 3 4B IT", "notes": "Official-recipe conversion; stock runtime.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "stock-runner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "4B", "precision": "int4", "quantization": "palettized" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": true, "upstream_repo": "google/gemma-3-4b-it" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-gemma-3-12b-it", "bundle_kind": "llm", "capabilities": [ "chat", "instruction-following", "text-generation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Gemma", "id": "official-gemma-3-12b-it", "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Gemma Terms" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Gemma 3 12B IT", "notes": "Official-recipe conversion; stock runtime.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "stock-runner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "12B", "precision": "int4", "quantization": "palettized" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": true, "upstream_repo": "google/gemma-3-12b-it" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-mistral-7b-v0-3", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Mistral", "id": "official-mistral-7b-v0-3", "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Mistral 7B v0.3", "notes": "Official-recipe conversion; stock runtime.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "stock-runner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "7B", "precision": "int4", "quantization": "palettized" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": true, "upstream_repo": "mistralai/Mistral-7B-Instruct-v0.3" }, { "architecture": "diffusion", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-flux-2-klein-4b", "bundle_kind": "image-generation", "capabilities": [ "image-generation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "FLUX", "id": "official-flux-2-klein-4b", "io_contract": { "entrypoint": { "framework": "CoreAIDiffusionPipeline", "init_pattern": "let pipeline = try await Flux2Pipeline(from: bundleURL, config: configSource, mode: decodeResolution)", "type": "Flux2Pipeline" }, "files": { "processor_ref": "pipeline.json", "tokenizer_ref": "tokenizer" }, "inputs": [ { "modality": "text", "name": "prompt", "swift_type": "String" } ], "outputs": [ { "decoding": { "format_selector": "1024x1024 default (pipeline.json image_size); default_steps 4, guidance 1.0" }, "name": "images", "swift_type": "[CGImage]" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "image" ] }, "name": "FLUX.2 klein 4B", "notes": "Official diffusion export; stock CoreAIDiffusionPipeline.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIDiffusionPipeline", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "4GB", "parameters": "4B / ~4GB", "precision": "int4", "quantization": "int4-per-block" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": false, "upstream_repo": "black-forest-labs/FLUX.2-klein-4B" }, { "architecture": "cnn/transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-sam-3", "bundle_kind": "segmentation", "capabilities": [ "promptable-segmentation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "SAM", "id": "official-sam-3", "io_contract": { "entrypoint": { "framework": "CoreAIImageSegmenter", "init_pattern": "let segmenter = try await ImageSegmenter(resourcesAt: bundleDir)", "type": "ImageSegmenter" }, "files": { "tokenizer_ref": "tokenizer" }, "inputs": [ { "modality": "image", "name": "image", "preprocessing": { "resize": "any size; the engine resizes internally" }, "swift_type": "CGImage" }, { "constraints": { "max_context": 32 }, "modality": "text", "name": "prompt", "swift_type": "String" } ], "outputs": [ { "decoding": { "coordinate_convention": "masks decoded to input-image pixel space (SegmentationPostprocessor.decode)", "score_range": "segments sorted by score descending" }, "name": "segments", "swift_type": "SegmentationResponse" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "check_license", "name": "Meta SAM License" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "masks", "boxes", "scores" ] }, "name": "SAM 3", "notes": "Promptable open-vocabulary segmentation; stock CoreAIImageSegmenter.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIImageSegmenter", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "1.7GB", "parameters": "~1.7GB", "precision": "fp16", "quantization": "none" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-whisper-large-v3-turbo", "bundle_kind": "asr", "capabilities": [ "speech-to-text" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Whisper", "id": "official-whisper-large-v3-turbo", "io_contract": { "entrypoint": { "framework": "CoreAI", "init_pattern": "let model = try await AIModel(contentsOf: aimodelURL)", "type": "AIModel" }, "files": { "processor_ref": "preprocessor_config.json", "tokenizer_ref": "tokenizer" }, "inputs": [ { "constraints": { "max_audio_s": 30 }, "modality": "audio", "name": "audio", "preprocessing": { "normalization": "WhisperFeatureExtractor log-mel: 128 mel bins, n_fft 400, hop_length 160", "sample_rate": 16000 } } ], "outputs": [ { "decoding": { "detokenization": "WhisperTokenizer (tokenizer/); autoregressive decode in a fixed 128-token window" }, "name": "transcript", "swift_type": "String" } ] }, "last_verified": "2026-06-24", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "audio" ], "output": [ "transcript" ] }, "name": "Whisper large-v3-turbo", "notes": "Bundle includes tokenizer and mel filterbank.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAITranscribe", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "1.5GB", "parameters": "809M / ~1.5GB", "precision": "fp16", "quantization": "none" }, "source_group": "official", "source_path": "https://github.com/john-rocky/coreai-model-zoo/tree/main/official", "sources": [ "coreai-model-zoo-official", "coreai-model-zoo-official-readme" ], "status": "confirmed", "streaming": false, "upstream_repo": "openai/whisper-large-v3-turbo" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "minicpm5-1b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "reasoning" ], "confidence": "medium", "context_window": "128K", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "MiniCPM", "id": "minicpm5-1b", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "MiniCPM5-1B", "notes": "Hybrid Think/No-Think reasoning; 128K context; int8 lossless (24/24 exact vs HF).", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "1.0GB", "parameters": "1.08B", "precision": "int8", "quantization": "int8" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/minicpm5-1b.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "openbmb/MiniCPM5-1B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "fastcontext-1-0-4b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "agentic" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "FastContext", "id": "fastcontext-1-0-4b", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "FastContext-1.0-4B", "notes": "Repository-exploration agent (Microsoft); Qwen3-4B arch; AOT h18p; ANE unsupported.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "2.1GB", "parameters": "4B", "precision": "int4", "quantization": "int4" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/fastcontext.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "microsoft/FastContext-1.0-4B-SFT" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "bitcpm-8b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "BitCPM", "id": "bitcpm-8b", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "BitCPM-8B", "notes": "First 1.58-bit ternary LLM; custom 2-bit packed-GEMM Metal kernel; ~2.1 GB resident on iPhone.", "runtime": { "aot_required": true, "custom_kernel": true, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "2.1GB", "parameters": "8B / ternary {-1,0,+1}", "precision": "1.58-bit ternary", "quantization": "ternary-TQ2.0" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/bitcpm-8b.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "openbmb/BitCPM-CANN-8B" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "holo2-4b", "bundle_kind": "vlm", "capabilities": [ "vision-language", "gui-grounding" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Holo", "id": "holo2-4b", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "coordinates" ] }, "name": "Holo2-4B", "notes": "GUI-grounding / computer-use VLM (H Company); Qwen3-VL-4B backbone; screenshot → click coordinates.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "4B", "precision": "int8", "quantization": "int8" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/holo2.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": true, "upstream_repo": "Hcompany/Holo2-4B" }, { "architecture": "encoder", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "colmodernvbert", "bundle_kind": "embedding", "capabilities": [ "visual-document-retrieval" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "ModernVBERT", "id": "colmodernvbert", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "vector" ] }, "name": "ColModernVBERT", "notes": "Visual document retriever; ColBERT/MaxSim late-interaction; ModernBERT + SigLIP2.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIKit-GraphModel", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "705MB", "parameters": "250M", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/colmodernvbert.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false, "upstream_repo": "ModernVBERT/colmodernvbert" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-asr-1-7b", "bundle_kind": "asr", "capabilities": [ "speech-to-text" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen3-asr-1-7b", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "audio" ], "output": [ "transcript" ] }, "name": "Qwen3-ASR-1.7B", "notes": "Zoo's first ASR; AuT encoder + Qwen3 decoder; 52 languages.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": false }, "size": { "artifact_size": "not_published", "parameters": "1.7B", "precision": "int8", "quantization": "int8lin" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/README.md", "sources": [ "coreai-model-zoo-readme" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3-ASR-1.7B" }, { "architecture": "encoder", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "parakeet-tdt-0-6b", "bundle_kind": "asr", "capabilities": [ "speech-to-text" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Parakeet", "id": "parakeet-tdt-0-6b", "last_verified": "2026-06-29", "license": { "commercial_use": "check_license", "name": "CC-BY-4.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "audio" ], "output": [ "transcript" ] }, "name": "Parakeet-TDT-0.6B", "notes": "First transducer/TDT (RNN-T); NVIDIA FastConformer + LSTM predictor; 25 EU languages; iPhone 47.9× realtime.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIKit-GraphModel", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "1.2GB", "parameters": "0.6B", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/parakeet.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "voxcpm2-2b", "bundle_kind": "tts", "capabilities": [ "text-to-speech" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "VoxCPM", "id": "voxcpm2-2b", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "audio" ] }, "name": "VoxCPM2 2B", "notes": "2B successor at 48 kHz; MiniCPM4 28L LM + LocDiT-12L flow-matching + 48 kHz AudioVAE.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "4.9GB", "parameters": "2B", "precision": "int8", "quantization": "int8" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/README.md", "sources": [ "coreai-model-zoo-readme" ], "status": "confirmed", "streaming": true }, { "architecture": "encoder", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "yolox-s", "bundle_kind": "object-detection", "capabilities": [ "object-detection" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "YOLOX", "id": "yolox-s", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "boxes", "classes", "scores" ] }, "name": "YOLOX-S", "notes": "First single-stage anchor-free detector (YOLO family); CSPDarknet + PAFPN; per-class NMS; 208 FPS on M4 Max GPU.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIKit-GraphModel", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "36.2MB", "parameters": "8.97M", "precision": "fp32", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/yolox.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "tripo-splat", "bundle_kind": "3d", "capabilities": [ "image-to-3d" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "TripoSplat", "id": "tripo-splat", "last_verified": "2026-06-29", "license": { "commercial_use": "check_license", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "gaussian-splats" ] }, "name": "TripoSplat", "notes": "Single image → 3D Gaussian splats; DINOv3 ViT-H + 20-step flow-matching DiT; .ply/.splat → RealityKit / MetalSplatter; VAST; Mac-only.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": false }, "size": { "artifact_size": "not_published", "parameters": "not_published", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/README.md", "sources": [ "coreai-model-zoo-readme" ], "status": "confirmed", "streaming": false }, { "architecture": "diffusion", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "ltx-video-2b", "bundle_kind": "video", "capabilities": [ "text-to-video" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "LTX-Video", "id": "ltx-video-2b", "last_verified": "2026-06-29", "license": { "commercial_use": "check_license", "name": "OpenRAIL-M" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "video" ] }, "name": "LTX-Video 2B distilled", "notes": "First video model; T5-XXL + 8-step flow-matching DiT + causal video VAE; 512×768×49f ~14s Mac GPU; Lightricks; Mac-only.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIVideoPipeline", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "13.5GB", "parameters": "2B / DiT 1.9B + T5-XXL 4.76B", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/ltxvideo.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-zoo" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-qwen3-1-7b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "official-qwen3-1-7b", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Qwen3 1.7B", "notes": "Fills the missing 1.7B rung in the dense Qwen3 Core AI line. GPU-only — ANE export omitted (loads but fails on iOS 27).", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "stock-runner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "1.7B", "precision": "int4", "quantization": "palettized" }, "source_group": "official", "source_path": "https://github.com/apple/coreai-models", "sources": [ "apple-coreai-models", "huggingface-mlboydaisuke" ], "status": "confirmed", "streaming": true, "upstream_repo": "Qwen/Qwen3-1.7B" }, { "architecture": "encoder", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "official-clip-vit-base-patch32", "bundle_kind": "embedding", "capabilities": [ "image-text-similarity" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "CLIP", "id": "official-clip-vit-base-patch32", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "vector" ] }, "name": "CLIP ViT-B/32", "notes": "Official recipe CLIP export (models/clip/export.py); fp16 static; text padded to 77 tokens for free-text queries. Image+text → cosine similarity.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIKit-GraphModel", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "151M", "precision": "fp16", "quantization": "none" }, "source_group": "official", "source_path": "https://github.com/apple/coreai-models", "sources": [ "apple-coreai-models", "huggingface-mlboydaisuke" ], "status": "confirmed", "streaming": false, "upstream_repo": "openai/clip-vit-base-patch32" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "llada-8b-dllm", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "diffusion-lm" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "LLaDA", "id": "llada-8b-dllm", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "LLaDA-8B dLLM", "notes": "First diffusion LM (dLLM) on Core AI. Masked bidirectional text generation, no KV cache. int4 viable (not a cliff). ~38 tok/s at threshold 1.0.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "4.9GB", "parameters": "8B", "precision": "int4", "quantization": "int4-per-block-32" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/knowledge/diffusion-llms-dllm.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-knowledge" ], "status": "confirmed", "streaming": true, "upstream_repo": "d3LLM/d3LLM_LLaDA" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "bitvla", "bundle_kind": "action", "capabilities": [ "vision-language-action", "robotics" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "BitVLA", "id": "bitvla", "last_verified": "2026-06-29", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image", "text" ], "output": [ "action-tokens" ] }, "name": "BitVLA", "notes": "First VLA / robotics model on Core AI. 1.58-bit ternary VLA: BitNet LLM + BitSigLIP vision → 7-DoF robot action. Generalized ternary kernel (arbitrary K/N).", "runtime": { "aot_required": false, "custom_kernel": true, "patch_required": true, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "2B (BitNet b1.58)", "precision": "1.58-bit ternary", "quantization": "ternary-W1.58-A8" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/knowledge/bitvla-1.58bit-vla.md", "sources": [ "coreai-model-zoo-readme", "coreai-model-zoo-knowledge" ], "status": "confirmed", "streaming": false }, { "architecture": "diffusion", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "stable-audio-open-small", "bundle_kind": "audio-generation", "capabilities": [ "text-to-audio", "music-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Stable Audio", "id": "stable-audio-open-small", "last_verified": "2026-06-29", "license": { "commercial_use": "check_license", "name": "Stability Community" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "audio" ] }, "name": "Stable Audio Open Small", "notes": "First generative audio model. Latent diffusion: T5 encoder + DiT (8-step rectified-flow) + Oobleck VAE. ~11s 44.1 kHz stereo in ~0.4s on M4 Max (≈30× real-time).", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIDiffusionPipeline", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "1GB", "parameters": "not_published", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/README.md", "sources": [ "coreai-model-zoo-readme" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "whisper-large-v3-turbo-carstenl", "bundle_kind": "asr", "capabilities": [ "speech-to-text" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": "unknown", "mac": true, "mac_only": false }, "family": "Whisper", "id": "whisper-large-v3-turbo-carstenl", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "active", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "audio" ], "output": [ "transcript" ] }, "name": "Whisper large-v3-turbo (CarstenL)", "notes": "Independent conversion with split encoder/decoder .aimodel bundles. 99 languages.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAITranscribe", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "3.1GB", "parameters": "809M", "precision": "fp16", "quantization": "none" }, "source_group": "external", "source_path": "https://huggingface.co/CarstenL/whisper-large-v3-turbo-coreai", "sources": [ "carstenl-whisper-coreai" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "whisper-tiny-en-intiser", "bundle_kind": "asr", "capabilities": [ "speech-to-text" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": "unknown", "mac": true, "mac_only": false }, "family": "Whisper", "id": "whisper-tiny-en-intiser", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "active", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "audio" ], "output": [ "transcript" ] }, "name": "Whisper tiny.en (Intiser)", "notes": "English-only Whisper tiny variant. Minimal footprint ASR.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAITranscribe", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "not_published", "parameters": "39M", "precision": "fp16", "quantization": "none" }, "source_group": "external", "source_path": "https://huggingface.co/Intiser/whisper-tiny-en-coreai", "sources": [ "intiser-whisper-coreai" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "whisper-medium-intiser", "bundle_kind": "asr", "capabilities": [ "speech-to-text" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": "unknown", "mac": true, "mac_only": false }, "family": "Whisper", "id": "whisper-medium-intiser", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "active", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "audio" ], "output": [ "transcript" ] }, "name": "Whisper medium (Intiser)", "notes": "Multilingual Whisper medium variant.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAITranscribe", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "not_published", "parameters": "769M", "precision": "fp16", "quantization": "none" }, "source_group": "external", "source_path": "https://huggingface.co/Intiser/whisper-medium-coreai", "sources": [ "intiser-whisper-coreai" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "gemma-4-12b-mm-warshanks", "bundle_kind": "vlm", "capabilities": [ "chat", "vision-language", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Gemma", "id": "gemma-4-12b-mm-warshanks", "last_verified": "2026-06-30", "license": { "commercial_use": "check_license", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text", "image", "audio" ], "output": [ "text" ] }, "name": "Gemma 4 12B IT Multimodal (warshanks)", "notes": "First multimodal Gemma 4 conversion: text+image+audio input via encoder-free architecture. Custom conversion recipe extending coreai-models. INT4 block-32 weight-only decoder. Multi-function bundle (main, prefill_multimodal, embed_vision, embed_audio).", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "6.4GB", "parameters": "12B", "precision": "int4", "quantization": "int4linsym" }, "source_group": "external", "source_path": "https://huggingface.co/warshanks/gemma-4-12B-it-coreai", "sources": [ "warshanks-gemma-4-12b-coreai" ], "status": "confirmed", "streaming": true, "upstream_repo": "google/gemma-4-12B-it" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "gemma-4-12b-assistant-warshanks", "bundle_kind": "llm", "capabilities": [ "speculative-decoding" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Gemma", "id": "gemma-4-12b-assistant-warshanks", "last_verified": "2026-06-30", "license": { "commercial_use": "check_license", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Gemma 4 12B IT Assistant (MTP draft)", "notes": "Multi-token-prediction draft model for speculative decoding with Gemma 4 12B. Cross-attends main model KV cache (layers 46/47) zero-copy. ~3ms/step on M4 Max vs ~21ms for the 12B. Companion to warshanks/gemma-4-12B-it-coreai.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "12B", "precision": "int4", "quantization": "int4linsym" }, "source_group": "external", "source_path": "https://huggingface.co/warshanks/gemma-4-12B-it-assistant-coreai", "sources": [ "warshanks-gemma-4-12b-coreai" ], "status": "confirmed", "streaming": false, "upstream_repo": "google/gemma-4-12B-it-assistant" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "vibevoice-0-5b", "bundle_kind": "tts", "capabilities": [ "text-to-speech" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "VibeVoice", "id": "vibevoice-0-5b", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "audio" ] }, "name": "VibeVoice 0.5B", "notes": "Streaming TTS with 0.5B Qwen2 backbone. 10.2x RTF on M4 Max. Multi-bundle pipeline: LM + diffusion sampler + acoustic decoder/connector. Microsoft VibeVoice original weights.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "0.5B", "precision": "fp16", "quantization": "int4" }, "source_group": "external", "source_path": "https://huggingface.co/gafiatulin/vibevoice-0.5b-coreai", "sources": [ "gafiatulin-vibevoice-coreai" ], "status": "confirmed", "streaming": true }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "vibevoice-1-5b", "bundle_kind": "tts", "capabilities": [ "text-to-speech" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "VibeVoice", "id": "vibevoice-1-5b", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "audio" ] }, "name": "VibeVoice 1.5B", "notes": "Multi-speaker TTS with 1.5B Qwen2 backbone. INT8 LM. 4.99x RTF on M4 Max.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "1.5B", "precision": "fp16", "quantization": "int8" }, "source_group": "external", "source_path": "https://huggingface.co/gafiatulin/vibevoice-1.5b-coreai", "sources": [ "gafiatulin-vibevoice-coreai" ], "status": "confirmed", "streaming": true }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "vibevoice-7b", "bundle_kind": "tts", "capabilities": [ "text-to-speech" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "VibeVoice", "id": "vibevoice-7b", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "audio" ] }, "name": "VibeVoice 7B", "notes": "High-quality multi-speaker TTS with 7B Qwen2 backbone. INT8/4 LM. 2.37x RTF on M4 Max.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "7B", "precision": "fp16", "quantization": "int8" }, "source_group": "external", "source_path": "https://huggingface.co/gafiatulin/vibevoice-7b-coreai", "sources": [ "gafiatulin-vibevoice-coreai" ], "status": "confirmed", "streaming": true }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "vibevoice-asr", "bundle_kind": "asr", "capabilities": [ "speech-to-text" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "VibeVoice", "id": "vibevoice-asr", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "audio" ], "output": [ "transcript" ] }, "name": "VibeVoice ASR", "notes": "On-device transcription, Qwen2-7B. INT8/4 LM. 11.1x RTF on M4 Max. Microsoft VibeVoice ASR original weights.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "7B", "precision": "fp16", "quantization": "int8" }, "source_group": "external", "source_path": "https://huggingface.co/gafiatulin/vibevoice-asr-coreai", "sources": [ "gafiatulin-vibevoice-coreai" ], "status": "confirmed", "streaming": true }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "nemotron-3-5-asr-streaming", "bundle_kind": "asr", "capabilities": [ "speech-to-text" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Nemotron", "id": "nemotron-3-5-asr-streaming", "last_verified": "2026-06-30", "license": { "commercial_use": "check_license", "name": "OpenMDW-1.1" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "audio" ], "output": [ "transcript" ] }, "name": "Nemotron 3.5 ASR Streaming 0.6B", "notes": "NVIDIA FastConformer + RNN-T streaming ASR. Cache-aware 320ms encoder mode. AOT artifacts for iOS (A19 h18p) + macOS (M3-M5). Companion Swift package mweinbach/NemotronCoreAI. Multiple quantization variants (int8, lut6, lut8). 174 files.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAITranscribe", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "not_published", "parameters": "0.6B", "precision": "fp16", "quantization": "none" }, "source_group": "external", "source_path": "https://huggingface.co/mweinbach/nemotron-3.5-asr-streaming-0.6b-coreai", "sources": [ "mweinbach-nemotron-coreai" ], "status": "confirmed", "streaming": false }, { "architecture": "diffusion", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "z-image-turbo", "bundle_kind": "image-generation", "capabilities": [ "image-generation" ], "confidence": "medium", "device_support": { "ipad": true, "iphone": true, "mac": true, "mac_only": false }, "family": "Z-Image", "id": "z-image-turbo", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "image" ] }, "name": "Z-Image Turbo", "notes": "Tongyi-MAI Z-Image Turbo text-to-image. GPU (fast-512) and ANE (ane-1024, hoisted-adaLN) bundles at 512 and 1024 resolution. For CoreAIStudio on macOS/iOS/iPadOS/visionOS.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIDiffusionPipeline", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "not_published", "precision": "fp16", "quantization": "none" }, "source_group": "external", "source_path": "https://huggingface.co/bryanbblewis11/Z-Image-Turbo-CoreAI", "sources": [ "bryanbblewis11-coreai" ], "status": "confirmed", "streaming": false, "upstream_repo": "Tongyi-MAI/Z-Image-Turbo" }, { "architecture": "cnn/transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "real-esrgan-x4v3", "bundle_kind": "super-resolution", "capabilities": [ "super-resolution" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": "unknown", "mac": true, "mac_only": false }, "family": "Real-ESRGAN", "id": "real-esrgan-x4v3", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "BSD-3-Clause" }, "maturity": "active", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "image" ] }, "name": "Real-ESRGAN x4v3", "notes": "4x super-resolution (SRVGGNetCompact). fp16, 256 tile size. For CoreAIStudio.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIImageSegmenter", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "not_published", "parameters": "not_published", "precision": "fp16", "quantization": "none" }, "source_group": "external", "source_path": "https://huggingface.co/bryanbblewis11/RealESRGAN-x4v3-CoreAI", "sources": [ "bryanbblewis11-coreai" ], "status": "confirmed", "streaming": false }, { "architecture": "cnn/transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "efficientsam3-tinyvit", "bundle_kind": "segmentation", "capabilities": [ "promptable-segmentation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "SAM", "id": "efficientsam3-tinyvit", "last_verified": "2026-06-30", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "active", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "image" ], "output": [ "masks" ] }, "name": "EfficientSAM3-TinyViT", "notes": "Lightweight promptable segmentation. Hosted by Lenitas app distribution. 392MB total with tokenizer. SHA-256 verified.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": true, "runner": "CoreAIImageSegmenter", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "374MB", "parameters": "not_published", "precision": "fp16", "quantization": "none" }, "source_group": "external", "source_path": "https://huggingface.co/lenitas/coreai-artifacts", "sources": [ "lenitas-coreai-artifacts" ], "status": "confirmed", "streaming": false }, { "architecture": "recurrent", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "rwkv7-goose-1-5b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "medium", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "RWKV", "id": "rwkv7-goose-1-5b", "last_verified": "2026-07-01", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "RWKV-7 Goose 1.5B", "notes": "First pure-recurrent / linear-attention LLM on Core AI. No attention, no KV cache — O(1) per-token decode with constant memory and unbounded context. WKV7 delta-rule matrix-state time-mix + sqrelu channel-mix. int8 weight-only quant (FFN only; recurrence projections kept fp16).", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "1.5B", "precision": "int8", "quantization": "int8-per-block-32" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/conversion/export_rwkv7_decode.py", "sources": [ "coreai-model-zoo-readme", "rwkv-upstream" ], "status": "confirmed", "streaming": true }, { "architecture": "encoder", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "art-vjepa2-vitl-ssv2", "bundle_kind": "video", "capabilities": [ "video-classification" ], "confidence": "low", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "V-JEPA", "id": "vjepa2-vitl-ssv2", "io_contract": { "entrypoint": { "framework": "CoreAI", "init_pattern": "let model = try await AIModel(contentsOf: aimodelURL)", "type": "AIModel" }, "inputs": [ { "modality": "video", "name": "pixel_values_videos", "preprocessing": { "color_format": "RGB", "normalization": "RGB 0..1 -> (x - mean)/std, mean=[0.485,0.456,0.406], std=[0.229,0.224,0.225]", "resize": "16 frames uniformly sampled, 256x256 center crop" }, "tensor": { "dtype": "float16", "layout": "B,F,C,H,W", "shape": [ 1, 16, 3, 256, 256 ] } } ], "outputs": [ { "decoding": { "label_vocab_ref": "labels.json (Something-Something v2, 174 classes)" }, "name": "logits" } ] }, "last_verified": "2026-07-02", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "video" ], "output": [ "action-tokens" ] }, "name": "V-JEPA 2 ViT-L (SSv2)", "notes": "First video world-model on Core AI. JEPA architecture (joint-embedding predictive) from Meta — learns video representations by predicting masked spatiotemporal patches. fp16, AOT h18p specialized. Trained on Something-Something v2 (action recognition). ViT-L backbone.", "runtime": { "aot_required": true, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIKit-GraphModel", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": false }, "size": { "artifact_size": "700MB", "parameters": "300M", "precision": "fp16", "quantization": "none" }, "source_group": "zoo", "source_path": "https://huggingface.co/mlboydaisuke/VJEPA2-ViTL-SSv2-CoreAI", "sources": [ "coreai-model-zoo-readme", "vjepa2-upstream" ], "status": "unknown", "streaming": false, "upstream_repo": "facebook/vjepa2-vitl-fpc16-256-ssv2" }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "ornith-1-0-9b", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation", "code-generation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": false, "mac": true, "mac_only": true }, "family": "Qwen", "id": "ornith-1-0-9b", "last_verified": "2026-07-03", "license": { "commercial_use": "likely", "name": "MIT" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Ornith 1.0 9B", "notes": "First agentic-coding model in the ecosystem. Qwen3.5 hybrid decoder (32 layers, 3:1 GatedDeltaNet/attention interleave) trained by DeepReinforce to jointly solve coding tasks and construct orchestration scaffolds. int8hu ship recipe delivers 48.3 tok/s decode on M4 Max. int4lin variant reaches 58.9 tok/s (+22%), 7.5 GB. 24/24 exact match vs fp32 oracle on both quantizations. Dense MLP(12288), hidden 4096, untied 248320-vocab lm_head. Max context 8192.", "runtime": { "aot_required": false, "custom_kernel": true, "patch_required": true, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": false, "tokenizer_required": true }, "size": { "artifact_size": "9.8GB", "parameters": "9B", "precision": "int8", "quantization": "int8hu" }, "source_group": "zoo", "source_path": "https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/ornith-1.0-9b.md", "sources": [ "coreai-model-zoo-readme", "ornith-upstream" ], "status": "confirmed", "streaming": false }, { "architecture": "transformer", "artifact": { "availability": "available", "format": "aimodel" }, "artifact_ref": "qwen3-enhancer", "bundle_kind": "llm", "capabilities": [ "chat", "text-generation" ], "confidence": "high", "device_support": { "ipad": "unknown", "iphone": true, "mac": true, "mac_only": false }, "family": "Qwen", "id": "qwen3-enhancer", "last_verified": "2026-07-04", "license": { "commercial_use": "likely", "name": "Apache-2.0" }, "maturity": "experimental", "min_os": { "ios": "27.0", "macos": "27.0" }, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "name": "Qwen3 Enhancer (Huihui Qwen3-4B Abliterated v2)", "notes": "Abliterated (uncensored) Qwen3-4B by huihui-ai. 4-bit dynamic quantization, 36 layers, hidden 2560, vocab 151936, max context 8192. Embedded tokenizer with chat template. Compiled 2026-06-20. Published by bryanbblewis11 outside the Zoo ecosystem.", "runtime": { "aot_required": false, "custom_kernel": false, "patch_required": false, "processor_required": false, "runner": "CoreAIRunner", "runtime_name": "apple-core-ai", "stock_runtime": true, "tokenizer_required": true }, "size": { "artifact_size": "not_published", "parameters": "4B", "precision": "int4", "quantization": "4bit-dynamic" }, "source_group": "external", "source_path": "https://huggingface.co/bryanbblewis11/Qwen3-Enhancer-CoreAI", "sources": [ "qwen3-enhancer-upstream" ], "status": "confirmed", "streaming": false, "upstream_repo": "huihui-ai/Huihui-Qwen3-4B-abliterated-v2" } ] }