{ "models": [ { "name": "Gemma-4-E2B-it", "modelId": "litert-community/gemma-4-E2B-it-litert-lm", "modelFile": "gemma-4-E2B-it.litertlm", "description": "A variant of Gemma 4 E2B ready for deployment on Android using [LiteRT-LM](https://github.com/google-ai-edge/LiteRT-LM/blob/main/docs/api/kotlin/getting_started.md). It supports multi-modality input, with up to 32K context length.", "sizeInBytes": 2588147712, "minDeviceMemoryInGb": 8, "commitHash": "6e5c4f1e395deb959c494953478fa5cec4b8008f", "llmSupportImage": true, "llmSupportAudio": true, "capabilities": ["llm_thinking", "speculative_decoding"], "defaultConfig": { "topK": 64, "topP": 0.95, "temperature": 1.0, "maxContextLength": 32000, "maxTokens": 4000, "accelerators": "gpu,cpu", "visionAccelerator": "gpu" }, "taskTypes": [ "llm_chat", "llm_prompt_lab", "llm_agent_chat", "llm_ask_image", "llm_ask_audio" ], "bestForTaskTypes": [ "llm_chat", "llm_prompt_lab", "llm_agent_chat", "llm_ask_image", "llm_ask_audio" ], "capabilityToTaskTypes": { "llm_thinking": ["llm_chat", "llm_ask_image", "llm_ask_audio"], "speculative_decoding": ["llm_chat", "llm_ask_image", "llm_ask_audio", "llm_prompt_lab"] }, "updatableModelFiles": [ { "fileName": "gemma-4-E2B-it.litertlm", "commitHash": "7fa1d78473894f7e736a21d920c3aa80f950c0db" } ], "updateInfo": "The updated model now leverages Gemma 4's built-in Multi Token Prediction (MTP) for faster decode on select workloads. You can enable or disable MTP using a new toggle in chat configuration." }, { "name": "Gemma-4-E4B-it", "modelId": "litert-community/gemma-4-E4B-it-litert-lm", "modelFile": "gemma-4-E4B-it.litertlm", "description": "A variant of Gemma 4 E4B ready for deployment on Android using [LiteRT-LM](https://github.com/google-ai-edge/LiteRT-LM/blob/main/docs/api/kotlin/getting_started.md). It supports multi-modality input, with up to 32K context length.", "sizeInBytes": 3659530240, "minDeviceMemoryInGb": 12, "commitHash": "28299f30ee4d43294517a4ac93abd6163412f07f", "llmSupportImage": true, "llmSupportAudio": true, "capabilities": ["llm_thinking", "speculative_decoding"], "defaultConfig": { "topK": 64, "topP": 0.95, "temperature": 1.0, "maxContextLength": 32000, "maxTokens": 4000, "accelerators": "gpu,cpu", "visionAccelerator": "gpu" }, "taskTypes": [ "llm_chat", "llm_prompt_lab", "llm_agent_chat", "llm_ask_image", "llm_ask_audio" ], "bestForTaskTypes": [ "llm_chat", "llm_prompt_lab", "llm_agent_chat", "llm_ask_image", "llm_ask_audio" ], "capabilityToTaskTypes": { "llm_thinking": ["llm_chat", "llm_ask_image", "llm_ask_audio"], "speculative_decoding": ["llm_chat", "llm_ask_image", "llm_ask_audio", "llm_prompt_lab"] }, "updatableModelFiles": [ { "fileName": "gemma-4-E4B-it.litertlm", "commitHash": "9695417f248178c63a9f318c6e0c56cb917cb837" } ], "updateInfo": "The updated model now leverages Gemma 4's built-in Multi Token Prediction (MTP) for faster decode on select workloads. You can enable or disable MTP using a new toggle in chat configuration." }, { "name": "EmbeddingGemma-2", "modelId": "litert-community/embeddinggemma-2-740m-litert-lm", "modelFile": "embeddinggemma-2-740m.litertlm", "description": "A variant of EmbeddingGemma 2 ready for deployment on Android using MediaPipe [UniversalEmbedder](https://developers.google.com/edge/mediapipe/solutions/retrieval/universal_embedder) and [SemanticRetriever](https://developers.google.com/edge/mediapipe/solutions/retrieval/semantic_retriever). It is a multimodal embedding model with image, audio, video, and text understanding.", "sizeInBytes": 484622336, "commitHash": "86b903a293f96d042bb92e1b7af333d94114458f", "taskTypes": ["smart_album", "video_moment_finder"], "bestForTaskTypes": ["smart_album", "video_moment_finder"], "extraDataFiles": [ { "name": "sample images", "url": "https://dl.google.com/google-ai-edge-gallery/android/smart_album/20260911/samples.zip", "downloadFileName": "samples.zip", "sizeInBytes": 15261357, "targetTaskTypes": ["smart_album"] } ], "defaultConfig": { "accelerators": "gpu,cpu" }, "variantLabel": "GPU" }, { "name": "Gemma-3n-E2B-it", "modelId": "google/gemma-3n-E2B-it-litert-lm", "modelFile": "gemma-3n-E2B-it-int4.litertlm", "description": "A variant of [Gemma 3n E2B](https://ai.google.dev/gemma/docs/gemma-3n) ready for deployment on Android using [LiteRT-LM](https://github.com/google-ai-edge/LiteRT-LM/blob/main/kotlin/README.md). It supports text, vision, and audio input, with 4096 context length.", "sizeInBytes": 3655827456, "minDeviceMemoryInGb": 8, "commitHash": "ba9ca88da013b537b6ed38108be609b8db1c3a16", "llmSupportImage": true, "llmSupportAudio": true, "defaultConfig": { "topK": 64, "topP": 0.95, "temperature": 1.0, "maxTokens": 4096, "accelerators": "cpu,gpu" }, "taskTypes": ["llm_chat", "llm_prompt_lab", "llm_ask_image", "llm_ask_audio"], "bestForTaskTypes": ["llm_ask_image", "llm_ask_audio"] }, { "name": "Gemma-3n-E4B-it", "modelId": "google/gemma-3n-E4B-it-litert-lm", "modelFile": "gemma-3n-E4B-it-int4.litertlm", "description": "A variant of [Gemma 3n E4B](https://ai.google.dev/gemma/docs/gemma-3n) ready for deployment on Android using [LiteRT-LM](https://github.com/google-ai-edge/LiteRT-LM/blob/main/kotlin/README.md). It supports text, vision, and audio input, with 4096 context length.", "sizeInBytes": 4919541760, "minDeviceMemoryInGb": 12, "commitHash": "297ed75955702dec3503e00c2c2ecbbf475300bc", "llmSupportImage": true, "llmSupportAudio": true, "defaultConfig": { "topK": 64, "topP": 0.95, "temperature": 1.0, "maxTokens": 4096, "accelerators": "cpu,gpu" }, "taskTypes": ["llm_chat", "llm_prompt_lab", "llm_ask_image", "llm_ask_audio"] }, { "name": "Gemma3-1B-IT", "modelId": "litert-community/Gemma3-1B-IT", "modelFile": "gemma3-1b-it-int4.litertlm", "description": "A variant of [google/Gemma-3-1B-IT](https://huggingface.co/google/Gemma-3-1B-IT) with 4-bit quantization ready for deployment on Android using [LiteRT-LM](https://github.com/google-ai-edge/LiteRT-LM/blob/main/kotlin/README.md).", "sizeInBytes": 584417280, "minDeviceMemoryInGb": 6, "commitHash": "42d538a932e8d5b12e6b3b455f5572560bd60b2c", "defaultConfig": { "topK": 64, "topP": 0.95, "temperature": 1.0, "maxTokens": 1024, "accelerators": "gpu,cpu" }, "taskTypes": ["llm_chat", "llm_prompt_lab"], "bestForTaskTypes": ["llm_chat", "llm_prompt_lab"] }, { "name": "Qwen2.5-1.5B-Instruct", "modelId": "litert-community/Qwen2.5-1.5B-Instruct", "modelFile": "Qwen2.5-1.5B-Instruct_multi-prefill-seq_q8_ekv4096.litertlm", "description": "A variant of [Qwen/Qwen2.5-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct) ready for deployment on Android using [LiteRT-LM](https://github.com/google-ai-edge/LiteRT-LM/blob/main/kotlin/README.md).", "sizeInBytes": 1597922856, "minDeviceMemoryInGb": 6, "commitHash": "5e6e7a58d03777898b31330a15d2e0a5a9107de3", "defaultConfig": { "topK": 20, "topP": 0.8, "temperature": 0.7, "maxTokens": 4096, "accelerators": "gpu,cpu" }, "taskTypes": ["llm_chat", "llm_prompt_lab"] }, { "name": "DeepSeek-R1-Distill-Qwen-1.5B", "modelId": "litert-community/DeepSeek-R1-Distill-Qwen-1.5B", "modelFile": "DeepSeek-R1-Distill-Qwen-1.5B_multi-prefill-seq_q8_ekv4096.litertlm", "description": "A variant of [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B) ready for deployment on Android using [LiteRT-LM](https://github.com/google-ai-edge/LiteRT-LM/blob/main/kotlin/README.md).", "sizeInBytes": 1833451520, "minDeviceMemoryInGb": 6, "commitHash": "e34bb88632342d1f9640bad579a45134eb1cf988", "defaultConfig": { "topK": 64, "topP": 0.95, "temperature": 1.0, "maxTokens": 4096, "accelerators": "gpu,cpu" }, "taskTypes": ["llm_chat", "llm_prompt_lab"] }, { "name": "TinyGarden-270M", "modelId": "litert-community/functiongemma-270m-ft-tiny-garden", "modelFile": "tiny_garden_q8_ekv1024.litertlm", "description": "Fine-tuned Function Gemma 270M model for Tiny Garden.", "sizeInBytes": 288964608, "minDeviceMemoryInGb": 6, "commitHash": "c205853ff82da86141a1105faa2344a8b176dfe7", "defaultConfig": { "topK": 64, "topP": 0.95, "temperature": 0.0, "maxTokens": 1024, "accelerators": "cpu" }, "taskTypes": [ "llm_tiny_garden" ], "bestForTaskTypes": [ "llm_tiny_garden" ] }, { "name": "MobileActions-270M", "modelId": "litert-community/functiongemma-270m-ft-mobile-actions", "modelFile": "mobile_actions_q8_ekv1024.litertlm", "description": "Fine-tuned Function Gemma 270M model for Mobile Actions.", "sizeInBytes": 288964608, "minDeviceMemoryInGb": 6, "commitHash": "38942192c9b723af836d489074823ff33d4a3e7a", "defaultConfig": { "topK": 64, "topP": 0.95, "temperature": 0.0, "maxTokens": 1024, "accelerators": "cpu" }, "taskTypes": [ "llm_mobile_actions" ], "bestForTaskTypes": [ "llm_mobile_actions" ] }, { "name": "Magic touch", "modelId": "magic_touch", "modelFile": "interactive_segmentation.task", "description": "This model identifies segments given image coordinates for an area of interest. The model uses a Convolutional Neural Network, similar to a MobileNetV3 architecture, with a customized decoder.", "url": "https://storage.googleapis.com/mediapipe-models/interactive_segmenter_v2/magic_touch/int8/latest/interactive_segmentation.task", "sizeInBytes": 30525312, "commitHash": "20260810", "taskTypes": ["mp_scrapbook"], "bestForTaskTypes": ["mp_scrapbook"] } ] }