aid: modal url: https://raw.githubusercontent.com/api-evangelist/modal-com/refs/heads/main/apis.yml apis: - aid: modal:modal-functions-api name: Modal Functions API tags: - Serverless - GPU - Functions - Python humanURL: https://modal.com/docs/guide/apps properties: - url: https://modal.com/docs/guide/apps type: Documentation - url: https://modal.com/docs/reference/modal.Function type: Documentation - url: https://modal.com/docs/reference/modal.App type: Documentation - url: openapi/modal-functions-openapi.yml type: OpenAPI - url: json-schema/modal-function-schema.json type: JSONSchema - url: json-ld/modal-com-context.jsonld type: JSONLD - type: NaftikoCapability url: capabilities/functions-functions.yaml description: Modal Functions are the core serverless primitive — Python functions backed by autoscaling container pools. They support optional GPU attachment (T4 through B200), Modal-built container images, Volumes, Secrets, retries, timeouts, scheduling, web endpoints, dynamic batching, and class-based stateful workers. Invoke synchronously (`.remote()`), asynchronously (`.spawn()`), or as a map (`.map()`). - aid: modal:modal-sandboxes-api name: Modal Sandboxes API tags: - Sandboxes - Agents - Code Execution - Isolation humanURL: https://modal.com/docs/guide/sandboxes properties: - url: https://modal.com/docs/guide/sandboxes type: Documentation - url: https://modal.com/docs/reference/modal.Sandbox type: Documentation - url: openapi/modal-sandboxes-openapi.yml type: OpenAPI - url: json-schema/modal-sandbox-schema.json type: JSONSchema - type: NaftikoCapability url: capabilities/sandboxes-sandboxes.yaml description: Modal Sandboxes are secure containers for executing untrusted user or agent code. Spawned at runtime with `Sandbox.create()`, they support arbitrary command execution, filesystem snapshots, port tunnels, volume mounts, GPU attachment, idle timeouts (up to 24h), readiness probes, named lookup, and tagging. Sandboxes are the workhorse primitive for LLM coding agents and rollout systems on Modal. - aid: modal:modal-volumes-api name: Modal Volumes API tags: - Storage - Distributed File System - Persistence humanURL: https://modal.com/docs/guide/volumes properties: - url: https://modal.com/docs/guide/volumes type: Documentation - url: https://modal.com/docs/reference/modal.Volume type: Documentation - url: openapi/modal-volumes-openapi.yml type: OpenAPI - type: NaftikoCapability url: capabilities/volumes-volumes.yaml description: Modal Volumes provide a high-performance distributed file system optimized for ML workflows. Volumes attach to Functions and Sandboxes via `Volume.from_name()`, support concurrent reads, explicit `commit()` and `reload()` semantics, sub-path mounting, and read-only mounts. Volumes v2 (beta) removes inode limits and supports hundreds of concurrent writers to distinct files. - aid: modal:modal-images-api name: Modal Images API tags: - Container Images - Build - Python humanURL: https://modal.com/docs/guide/images properties: - url: https://modal.com/docs/guide/images type: Documentation - url: https://modal.com/docs/reference/modal.Image type: Documentation - url: openapi/modal-images-openapi.yml type: OpenAPI - type: NaftikoCapability url: capabilities/images-images.yaml description: Modal Images are declarative, layered container image specifications built lazily on Modal's backend. Build from `debian_slim`, `micromamba`, `from_registry`, or `from_dockerfile`; layer `pip_install`, `apt_install`, `run_commands`, `env`, `copy_local_file`, and `copy_local_dir`. Modal caches layers globally so cold starts of identical images are sub-second. - aid: modal:modal-secrets-api name: Modal Secrets API tags: - Secrets - Configuration - Environment Variables humanURL: https://modal.com/docs/guide/secrets properties: - url: https://modal.com/docs/guide/secrets type: Documentation - url: https://modal.com/docs/reference/modal.Secret type: Documentation - url: openapi/modal-secrets-openapi.yml type: OpenAPI - type: NaftikoCapability url: capabilities/secrets-secrets.yaml description: Modal Secrets are encrypted, named bundles of environment variables injected into Functions and Sandboxes at runtime. Manage via dashboard, the `modal secret` CLI, or programmatically with `Secret.from_name()` / `Secret.from_dict()`. - aid: modal:modal-dicts-queues-api name: Modal Dicts and Queues API tags: - Dicts - Queues - State - Coordination humanURL: https://modal.com/docs/guide/dicts properties: - url: https://modal.com/docs/guide/dicts type: Documentation - url: https://modal.com/docs/guide/queues type: Documentation - url: https://modal.com/docs/reference/modal.Dict type: Documentation - url: https://modal.com/docs/reference/modal.Queue type: Documentation - url: openapi/modal-dicts-queues-openapi.yml type: OpenAPI - type: NaftikoCapability url: capabilities/dicts-queues.yaml description: Modal Dicts and Queues are distributed in-memory primitives. `modal.Dict` is a named key-value store; `modal.Queue` is a FIFO queue with optional partitioning and blocking gets. Both are addressable by name across Functions and Sandboxes for cross-container coordination. - aid: modal:modal-schedules-api name: Modal Schedules API tags: - Cron - Scheduling - Recurring Jobs humanURL: https://modal.com/docs/guide/cron properties: - url: https://modal.com/docs/guide/cron type: Documentation - url: openapi/modal-schedules-openapi.yml type: OpenAPI - type: NaftikoCapability url: capabilities/schedules-schedules.yaml description: Modal Schedules attach periodic execution to a Function via `@app.function(schedule=modal.Cron(...))` or `modal.Period(...)`. Cron supports full cron syntax with timezone; Period specifies a fixed interval. Past runs appear in the dashboard with manual "run now" support. - aid: modal:modal-web-endpoints-api name: Modal Web Endpoints API tags: - Web - HTTP - ASGI - WSGI - FastAPI humanURL: https://modal.com/docs/guide/webhooks properties: - url: https://modal.com/docs/guide/webhooks type: Documentation - url: https://modal.com/docs/guide/webhook-urls type: Documentation - url: https://modal.com/docs/guide/webhook-proxy-auth type: Documentation - url: openapi/modal-web-endpoints-openapi.yml type: OpenAPI - type: NaftikoCapability url: capabilities/web-endpoints.yaml description: Modal Web Endpoints expose Functions over HTTP via `@modal.fastapi_endpoint`, `@modal.asgi_app` (FastAPI/Starlette/FastHTML), `@modal.wsgi_app` (Flask/Django), or `@modal.web_server` (any port-binding server). Supports streaming, WebSockets, proxy-auth tokens, custom URL labels, and live development via `modal serve`. - aid: modal:modal-tokens-api name: Modal Tokens and Administration API tags: - Administration - Tokens - Workspaces - Environments humanURL: https://modal.com/docs/reference/cli/token properties: - url: https://modal.com/docs/reference/cli/token type: Documentation - url: https://modal.com/docs/reference/cli/profile type: Documentation - url: https://modal.com/docs/guide/environments type: Documentation - url: openapi/modal-tokens-openapi.yml type: OpenAPI - type: NaftikoCapability url: capabilities/tokens-administration.yaml description: Administrative surface for Modal — token id/secret pair creation and revocation, environment management (e.g. `dev`/`staging`/`prod`), and workspace metadata. Tokens authenticate the `modal` CLI and the Python/TypeScript/Go SDKs against the control plane at `api.modal.com`. name: Modal tags: - Serverless - GPU - Cloud Compute - AI Infrastructure - Sandboxes - Inference - Training - Batch Processing - Python - TypeScript - Go kind: contract image: https://kinlane-productions2.s3.amazonaws.com/apis-json/apis-json-logo.jpg access: 3rd-Party common: - type: Portal url: https://modal.com - type: Documentation url: https://modal.com/docs - type: GettingStarted url: https://modal.com/docs/guide - url: https://modal.com/docs/examples name: Modal Examples Gallery type: CodeExamples - url: https://modal.com/docs/reference name: Python SDK Reference type: Documentation - url: https://modal.com/blog type: Blog - url: https://modal.com/changelog type: ChangeLog - url: https://status.modal.com type: StatusPage - url: https://modal.com/legal/terms type: TermsOfService - url: https://modal.com/legal/privacy type: PrivacyPolicy - url: https://trust.modal.com type: TrustCenter - url: https://modal.com/signup type: SignUp - url: https://modal.com/login type: SignUp - url: https://modal.com/support type: Support - url: https://modal.com/slack name: Modal Slack Community type: Forum - url: https://twitter.com/modal_labs type: Twitter - url: https://www.linkedin.com/company/modal-labs type: LinkedIn - url: https://github.com/modal-labs type: GitHubOrganization - url: https://github.com/modal-labs/modal-client name: Modal Python SDK (modal-client) type: SDK - url: https://github.com/modal-labs/libmodal name: Modal JavaScript/TypeScript and Go SDKs (libmodal) type: SDK - url: https://github.com/modal-labs/modal-examples name: Modal Examples type: CodeExamples - url: https://github.com/modal-labs/multinode-training-guide name: Multi-Node Training Guide type: CodeExamples - url: https://github.com/modal-labs/awesome-modal name: Awesome Modal type: Documentation - url: https://github.com/modal-labs/synchronicity name: Synchronicity (async/sync interop) type: Tool - url: https://github.com/modal-labs/quillman name: QuiLLMan voice chat demo type: CodeExamples - url: https://github.com/modal-labs/turbo-art name: Turbo Art (SDXL Turbo demo) type: CodeExamples - url: https://github.com/modal-labs/ci-on-modal name: CI on Modal sample type: CodeExamples - url: https://github.com/modal-labs/vercel-integration-example name: Modal Vercel Integration example type: Integrations - url: https://github.com/modal-labs/credential-injection name: Credential injection recipe for Sandboxes type: CodeExamples - url: https://github.com/modal-labs/gpu-glossary name: Modal GPU Glossary type: Documentation - url: https://github.com/modal-labs/stopwatch name: Stopwatch LLM benchmarking tool type: Tool - url: https://pypi.org/project/modal/ name: modal on PyPI type: Package - url: https://www.npmjs.com/package/modal name: modal on npm type: Package - url: https://modal.com/docs/reference/cli/run name: modal CLI reference type: Tool - url: https://modal.com/docs/guide/gpu type: Documentation - url: https://modal.com/docs/guide/scale type: Documentation - url: https://modal.com/docs/guide/cold-start type: Documentation - url: https://modal.com/docs/guide/memory-snapshots type: Documentation - url: https://modal.com/docs/guide/retries type: Documentation - url: https://modal.com/docs/guide/timeouts type: Documentation - url: https://modal.com/docs/guide/preemption type: Documentation - url: https://modal.com/docs/guide/tunnels type: Documentation - url: https://modal.com/docs/guide/cloud-bucket-mounts type: Documentation - url: https://modal.com/docs/guide/notebooks type: Documentation - url: https://modal.com/docs/guide/private-vpc type: Documentation - url: https://modal.com/docs/guide/region-selection type: Documentation - url: https://modal.com/docs/guide/managing-deployments type: Documentation - url: https://modal.com/docs/guide/observability type: Documentation - url: https://modal.com/docs/guide/security type: Documentation - url: https://modal.com/pricing data: - id: starter name: Starter entries: - geo: US unit: 1 label: User limit: 1 price: Free metric: user timeFrame: month description: Free tier with $30 monthly compute credit. elements: - name: $30 monthly compute credit included - name: Up to 100 concurrent containers - name: Up to 10 concurrent GPUs - name: 5 deployed cron jobs - name: 8 deployed web endpoints description: Free for individual developers. - id: team name: Team entries: - geo: US unit: 1 label: Workspace limit: 1 price: 250 metric: workspace timeFrame: month description: Team tier with $100 monthly compute credit and higher quotas. elements: - name: $100 monthly compute credit included - name: Up to 1,000 concurrent containers - name: Up to 50 concurrent GPUs - name: Unlimited cron jobs and web endpoints - name: Team collaboration description: For teams running production workloads. - id: enterprise name: Enterprise entries: - geo: US unit: 1 label: Workspace limit: 1 price: Call metric: workspace timeFrame: month description: Custom-priced enterprise contracts with volume discounts and private support. elements: - name: Volume discounts on compute - name: Custom container and GPU concurrency limits - name: Embedded ML services - name: Private Slack support channel - name: Negotiable SLAs and security review description: For organizations operating at scale. name: Plans type: Plans - url: https://modal.com/pricing data: - id: cpu name: CPU type: Compute entries: - geo: US unit: 1 label: Per core per second price: 0.0000131 metric: core-second timeFrame: usage description: Per physical CPU core per second. Minimum 0.125 cores per container. description: Per-second CPU pricing for Modal Functions. - id: memory name: Memory type: Compute entries: - geo: US unit: 1 label: Per GiB per second price: 0.00000222 metric: gib-second timeFrame: usage description: Per-second memory pricing for Modal Functions. - id: gpu-t4 name: NVIDIA T4 type: GPU entries: - geo: US unit: 1 label: Per second price: 0.000164 metric: gpu-second timeFrame: usage description: 16GB Turing GPU for cost-sensitive inference. - id: gpu-l4 name: NVIDIA L4 type: GPU entries: - geo: US unit: 1 label: Per second price: 0.000222 metric: gpu-second timeFrame: usage description: 24GB Ada Lovelace GPU. - id: gpu-a10 name: NVIDIA A10 type: GPU entries: - geo: US unit: 1 label: Per second price: 0.000306 metric: gpu-second timeFrame: usage description: 24GB Ampere GPU. - id: gpu-l40s name: NVIDIA L40S type: GPU entries: - geo: US unit: 1 label: Per second price: 0.000542 metric: gpu-second timeFrame: usage description: 48GB Ada Lovelace GPU. - id: gpu-a100-40gb name: NVIDIA A100 40GB type: GPU entries: - geo: US unit: 1 label: Per second price: 0.000583 metric: gpu-second timeFrame: usage description: 40GB Ampere GPU. - id: gpu-a100-80gb name: NVIDIA A100 80GB type: GPU entries: - geo: US unit: 1 label: Per second price: 0.000694 metric: gpu-second timeFrame: usage description: 80GB Ampere GPU. - id: gpu-rtx-pro-6000 name: NVIDIA RTX PRO 6000 type: GPU entries: - geo: US unit: 1 label: Per second price: 0.000842 metric: gpu-second timeFrame: usage description: 96GB Ada Lovelace workstation GPU. - id: gpu-h100 name: NVIDIA H100 type: GPU entries: - geo: US unit: 1 label: Per second price: 0.001097 metric: gpu-second timeFrame: usage description: 80GB Hopper GPU for training and LLM inference. - id: gpu-h200 name: NVIDIA H200 type: GPU entries: - geo: US unit: 1 label: Per second price: 0.001261 metric: gpu-second timeFrame: usage description: 141GB Hopper GPU with HBM3e memory. - id: gpu-b200 name: NVIDIA B200 type: GPU entries: - geo: US unit: 1 label: Per second price: 0.001736 metric: gpu-second timeFrame: usage description: Blackwell GPU — Modal's most powerful tier. - id: storage name: Volume Storage type: Storage entries: - geo: US unit: 1 label: Per GiB per month price: 0.09 metric: gib-month timeFrame: month description: 1 TiB included free monthly. description: Modal Volume persistent storage. - id: sandbox-cpu name: Sandbox CPU type: Sandbox entries: - geo: US unit: 1 label: Per core per second price: 0.00003942 metric: core-second timeFrame: usage description: Per-second CPU pricing for Modal Sandboxes and Notebooks. - id: sandbox-memory name: Sandbox Memory type: Sandbox entries: - geo: US unit: 1 label: Per GiB per second price: 0.00000672 metric: gib-second timeFrame: usage description: Per-second memory pricing for Modal Sandboxes and Notebooks. name: Pricing type: Pricing - url: plans/modal-plans-pricing.yml type: Plans - url: rate-limits/modal-rate-limits.yml type: RateLimits - url: finops/modal-finops.yml type: FinOps - url: vocabulary/modal-com-vocabulary.yml type: Vocabulary - url: rules/modal-rules.yml type: SpectralRules - type: Features data: - Serverless Python functions with sub-second cold starts and autoscaling container pools - GPU compute across T4, L4, A10, L40S, A100 40/80GB, RTX PRO 6000, H100, H200, and B200 - Modal Sandboxes for executing untrusted user/agent code with full container isolation - Declarative Image builder with layered caching (pip_install, apt_install, run_commands, from_registry, from_dockerfile) - Modal Volumes — distributed file system with commit/reload semantics; v2 supports many concurrent writers - Cloud bucket mounts for S3, GCS, R2, and Azure - Dicts and Queues for cross-container shared state and coordination - Modal Secrets for encrypted environment-variable injection - Modal Cron and Period schedules with timezone support - Web Endpoints via FastAPI, ASGI, WSGI, and custom web servers — streaming and WebSockets - Proxy-auth tokens, custom URL labels, and live development with `modal serve` - `modal run`, `modal deploy`, `modal serve`, `modal shell`, `modal container exec` CLI workflows - Python SDK (`modal`) plus JavaScript/TypeScript and Go SDKs via `libmodal` - Modal Notebooks for interactive development with GPU attachment - Multi-node distributed training with gang scheduling - Memory snapshots and warm pools for sub-second cold starts on large models - Per-second metered billing for CPU, memory, and GPU with tag-based chargeback - Plan-based concurrency quotas (containers, GPUs, cron jobs, web endpoints) - Token id/secret authentication scoped to workspace and environment - Static outbound IP via `modal.Proxy` for IP-allowlisted integrations - Tunnels and port forwarding for sandbox networking - SOC 2 Type II certified sources: - https://modal.com - https://modal.com/docs - https://modal.com/docs/guide - https://modal.com/docs/reference - https://modal.com/pricing - https://github.com/modal-labs updated: '2026-05-25' created: '2026-05-25T00:00:00.000Z' modified: '2026-05-25' position: Consuming description: Modal is a serverless cloud platform for AI and data workloads. Modal lets developers write ordinary Python and run it on remote GPUs and CPUs with sub-second cold starts, instant autoscaling, and declarative container images. The platform's primitives — Functions, Sandboxes, Volumes, Images, Secrets, Dicts, Queues, Schedules, and Web Endpoints — cover inference, fine-tuning, multi-node training, batch processing, agent code execution, and HTTP/web APIs. Modal sells per-second metered compute across the full NVIDIA GPU lineup (T4 → B200) under Starter, Team, and Enterprise plans. Modal Labs raised a $355M Series C in 2026 and is used in production by Decagon, Runway, Physical Intelligence, Suno, Chai Discovery, Lovable, Quora, Reducto, and many others. maintainers: - FN: Kin Lane email: info@apievangelist.com X: apievangelist url: https://apievangelist.com specificationVersion: '0.16'