Modal

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
    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
    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
    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
    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
    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
    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
    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
    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
    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-images.s3.amazonaws.com/shared/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'