Modal
Also known as: Modal Labs
Serverless cloud for AI workloads, with per second GPU billing, isolated sandboxes for untrusted agent code, scheduled functions and model endpoints.
Modal is a serverless cloud for AI workloads, organized around inference, sandboxes and training. Developers define functions in Python, JavaScript or Go with the GPU, image, memory and concurrency they need, and Modal builds the container, schedules it, scales it from zero and bills per second.
For agents, Modal Sandboxes run untrusted code in isolated environments created programmatically, with networking controls and snapshots of filesystem and memory. VM Sandboxes, now generally available, give an agent a full Linux computer for Docker stacks, local databases, graphical environments and mobile simulators, and Sidecars, in beta, keep trusted code such as credentials and harness logic in a separate container beside the sandbox. Restricted Functions are a second way to run untrusted code. Scheduled Functions, web endpoints and job queues start work without a person, and retries, queues and fan out handle scale.
The Modal Library serves models on Shared Endpoints billed per token or Dedicated Endpoints with the customer's own capacity and region, any open model can be deployed on GPUs with vLLM or SGLang, and multi node clusters, the Dojo training library and the open source Spindle training server cover post training. Its examples include coding agents, a computer use agent over VNC and parallel evals. Customers it names include DoorDash, Cognition, Ramp, Linear, Hebbia and Runway. Cognition runs reinforcement learning sandboxes and production inference on Modal, Ramp built its Inspect tool on Modal Sandboxes, and in July 2026 Modal ran a million sandboxes at once, all created in under a minute.
Workloads can be pinned to a region, with a data residency guide and a price multiplier for pinning. Modal states a SOC 2 Type II audit and HIPAA support, and offers Okta, Entra and SAML SSO, SCIM, role based access control and, on Enterprise, audit logs, with logs and metrics exportable to OpenTelemetry providers and Datadog.
Pricing is per second usage on top of a plan. Starter is free with $30 a month of compute, Team is $250 a month with $100 of compute, and Enterprise is custom. Sandbox CPU costs about three times the function rate, and network egress beyond the plan's included amount bills at $0.04 per GiB.
From our comparison video
Daytona vs Modal: What Agent Sandboxes Cost on Each
Our 4 minute video puts Modal side by side with Daytona. It covers what each one does and how each is priced, then when to choose which.
Recorded in October 2026. The profile here is live and updates as research is verified.
Vendor details
Canonical URL
https://modal.com
Category
Agent infrastructure
Subcategory
Compute and sandboxes
Funding status
Independent.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Python, JavaScript/TypeScript and Go SDKs and a CLI; Functions, Sandboxes, Endpoints, Scheduled Functions, job queues, Volumes, Dicts and Queues; OIDC to external clouds, cloud bucket mounts, OpenTelemetry and Datadog export, Okta, Entra and SAML SSO, SCIM (beta) and Slack notifications (beta).
In practice
Your agent needs to run untrusted code it generated. Modal Sandboxes spin up isolated environments programmatically with any dependency and scale back to zero when done.
You want serverless GPUs without DevOps. You declare a GPU function in Python, and Modal builds the container, schedules it on an H100 and bills per second.
Your agent should check a data source every morning without anyone starting it. You deploy it as a Modal Scheduled Function that runs on a cron schedule and writes its results to a Volume.
Sources & related URLs
Agentic Index coverage score
8.0 / 14 capabilities · 57%
| Integrations & Tool Calling | Partial |
|---|---|
|
Code on Modal reaches other systems through what the customer writes. OIDC tokens authenticate functions to external clouds, cloud buckets mount as filesystems, and secrets hold credentials. Slack notifications report on apps (beta), and examples connect Discord, Google Sheets and Tailscale. Modal has no catalog of connectors, OAuth actions or tool gateway of its own, so the tools are the customer's code. Sourcemodal.com/llms.txtread 2026-09-22 |
|
| Workflow Orchestration | Partial |
|
Retries are first class on Modal. A Retries policy can sit on any function, alongside job queues, fan out with map and spawn, batch and dynamic batching, and gang scheduled clusters. A pipeline that mixes model calls with deterministic steps can be built on these pieces. Sequencing and branching live in the customer's Python. Modal has no multistep workflow primitive that records completed steps and resumes after failure. Sourcemodal.com/docs/guide/retriesread 2026-09-22 |
|
| Knowledge Grounding & RAG | Not documented |
|
Retrieval systems the customer builds can run on Modal, and its examples embed documents with TEI and run a RAG chatbot over PDFs. Modal does not maintain a retrieval structure over the customer's content that an agent queries. Volumes, Dicts and mounted buckets are storage, and the product itself cannot index, refresh or permission sources. Sourcemodal.com/llms.txtread 2026-09-22 |
|
| Human Oversight & Guardrails | Not documented |
|
Work running on Modal has no approval step, consent checkpoint or escalation to a person. Restricted Functions stop untrusted code from touching Modal resources, calling other functions or reaching Modal's APIs. With sandbox networking controls, they give isolation the developer configures, and no person reviews an agent's action. Sourcemodal.com/docs/guide/restricted-accessread 2026-09-22 |
|
| Security, Identity & Governance | Full |
|
A SOC 2 Type II audit report is available through Modal's Security Portal at trust.modal.com, and Modal supports HIPAA and has external penetration testing. Identity runs through SSO with Okta, Microsoft Entra or custom SAML, with SCIM provisioning (beta). Access is managed with role based access control, user groups and service users, and an append only audit log records sensitive actions on Enterprise. Secrets, Restricted Functions that cannot touch Modal resources, and sandbox networking controls round out the controls, and Sidecars (beta) keep credentials and harness logic in a separate container the agent's code cannot reach. Sourcemodal.com/docs/guide/securityread 2026-10-23 |
|
| Observability & Auditability | Partial |
|
The workload itself is recorded in detail. Modal keeps real time logs and metrics per function and container, GPU metrics, function stats, and an append only audit log of sensitive workspace actions on Enterprise. Audit logs, function logs and container metrics export to any OpenTelemetry provider or Datadog. Audit stays separate from runtime. Prompts, tool calls, retrieved knowledge and outputs can be inspected step by step only as far as the customer's own code logs them. There is no tracing of each call an agent makes. Sourcemodal.com/docs/guide/otel-integrationread 2026-09-22 |
|
| Memory & State Persistence | Partial |
|
State carries across runs. Sandbox snapshots save and restore a sandbox's filesystem and memory, memory snapshots speed cold starts, volumes persist files, and Dicts and Queues hold data shared across function calls. The agent's code reads that state to carry on. Restored process state belongs to the machine, and the customer can resume or delete it but cannot review or edit it. Modal has no memory layer for agents. Sourcemodal.com/docs/guide/sandbox-snapshotsread 2026-09-22 |
|
| Deployment & Data Residency | Full |
|
Workloads can be pinned to a region. Functions and Sandboxes take a region argument for the container region, dedicated Endpoints set both container and routing regions, and a data residency guide sets out per service what can be pinned and when customer data may leave the region. Region selection carries a price multiplier (1.15x broad, 1.75x narrow). Modal is a managed cloud with no self hosted option, and shared Endpoints do not take a region. Sourcemodal.com/docs/guide/data-residencyread 2026-09-22 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
Ready made starting points come as a gallery of runnable examples and a Library of models deployable as Shared or Dedicated Endpoints. The agent examples include Cursor cloud agents, a background coding agent with OpenCode, a coding platform, a Claude Agent SDK Slack bot, a LangGraph agent in a GPU sandbox, a computer use agent over VNC and parallel evals. Developers build from these examples and models, and there is no browsable catalog of agents for a customer to adopt. Sourcemodal.com/llms.txtread 2026-09-22 |
|
| Triggers & Channel Coverage | Full |
|
Scheduled Functions run on a cron expression or a fixed period. Web functions and endpoints boot a container from zero when a request arrives, and job queues hand work to functions as it lands. Sourcemodal.com/docs/guide/cronread 2026-09-22 |
|
| Model Flexibility & Routing | Full |
|
Model choice is the customer's throughout. Modal serves models from its Library on Shared Endpoints billed per token or Dedicated Endpoints with the customer's capacity and region. The customer can deploy any open model on GPUs with vLLM or SGLang, as its examples show for DeepSeek, Nemotron and others. Sourcemodal.com/docs/guide/shared-endpointsread 2026-09-22 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Official SDKs in Python, JavaScript/TypeScript and Go (the latter two in beta) let outside code call Modal, alongside a full CLI that includes a skills command for coding agents. Web functions and endpoints expose the customer's code as HTTP APIs, and a gRPC API sits behind the open source client. Sourcemodal.com/llms.txtread 2026-09-22 |
|
| Testing, Debugging & Optimization | Not documented |
|
Evaluations run on Modal, but scoring them is left to other tools. Its example runs massively parallel evals with Harbor in Modal Sandboxes, and its reinforcement learning examples use verl and TRL, with the datasets, graders and scoring belonging to those frameworks. Modal has no fixtures, scoring, quality gates or comparison of agent runs of its own. Sourcemodal.com/llms.txtread 2026-09-22 |
|
| Browser & Computer Use | Partial |
|
Computer use agents can run on Modal's machines. VM Sandboxes, now generally available, give an agent a full Linux computer that can run Docker stacks, graphical environments and mobile simulators. In one example, a Browser Use agent drives Chromium inside a Modal Sandbox, watched over VNC, with the model served from a Modal Endpoint. Control of the browser comes from Browser Use, a separate library. Modal has no API for screenshots, clicks or typing, so it supplies the environment but not the control. Sourcemodal.com/blog/vm-sandboxes-agent-computersread 2026-10-23 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Sandboxes can now save their filesystem automatically on exit with Exit Snapshots, available in beta. A snapshot is taken on graceful and ungraceful exits alike, including terminate(), entrypoint completion, idle timeout, lifetime expiry and out of memory errors.
Bears on: Memory / state
View sourceModal Sandboxes can now run on a full virtual machine instead of the gVisor container runtime, set with one option in the Python and JavaScript SDKs. Modal says the VM runtime handles workloads such as running Docker inside a sandbox, while gVisor stays the default for now.
Bears on: Deployment / data residency
View sourcePricing
Starter is free with $30 a month of compute, Team is $250 a month, and Enterprise is custom, with usage billed per second.
per second compute (GPU, CPU, memory)
Included quota
Starter: $30 a month of compute, 3 seats, 100 containers, 10 concurrent GPUs. Team at $250 a month: $100 a month of compute, unlimited seats, 5,000 containers, 50 concurrent GPUs. Usage is billed per second on top of the included compute.
What is public
Modal publishes its plans, per second GPU, CPU and memory rates, sandbox rates, Volume storage pricing, region multipliers and egress allowances by plan.
Billing mechanics
Active seconds multiplied by the per second rate, plus the plan fee. Pinning a region multiplies base prices by 1.15x to 1.75x, and sandboxes bill at the non preemptible rate of about 3x. Scale to zero means no charge during idle time. Egress beyond the included amount bills at $0.04 per GiB.
Cost watchouts
Sandbox CPU is billed at about three times the function rate, which compounds for agents that keep sandboxes alive. Pinning a region adds 1.15x to 1.75x, Volumes bill $0.09 per GiB-month after the first TiB, and network egress beyond the plan's included amount (1 TiB on Starter, 10 TiB on Team) bills at $0.04 per GiB.
Variable cost rationale
There is a small plan floor, but cost is almost entirely per second compute that scales with GPU, CPU, and memory usage, amplified by regional and non preemption multipliers on sandbox workloads.
Additional watchouts
Budget sandbox time at the higher sandbox CPU rate, check region multipliers before pinning a region, and watch egress once it passes the plan's included amount.
Overage / add-ons
Per second usage across GPU, CPU and memory accrues above the plan's included compute, with region multipliers where a region is pinned. Volumes and egress bill beyond their included amounts.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free Starter tier with $30 a month in credits
Lowest paid plan
Team $250 a month plus per second usage
Commercial notes
Modal says it crossed $1 billion in annualized run rate with its Series D in October 2026. Customers it names include DoorDash, Cognition, Ramp, Linear, Hebbia and Runway, and graduate students and labs can apply for up to $10,000 in free compute credits.
Key ambiguities
Total cost depends on GPU type, CPU and memory time, sandbox use, region pinning, storage beyond 1 TiB and egress beyond the plan's allowance; Enterprise terms are custom.
Missing data
Enterprise pricing is custom.
Related vendors
- AgentOps — Agent observability and debugging platform: open source SDKs trace…
- AgentPhone — AgentPhone gives the AI agents a team builds their own US or…
- Agno — Python agent framework and AgentOS runtime (formerly Phidata) for…
- AIsa — Resource and payment gateway for AI agents, with one key to 110+…
- Albato — Albato is a no code integration platform that sells an embedded,…
- AlphaBitCore — AI control plane for regulated financial firms: one gateway enforces…
Alternatives to Modal
The closest documented capability profiles to Modal among agent infrastructure platforms tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Beam7.5 / 14Fuller documented coverage on Workflow OrchestrationModal vs Beam →
- Exa6.5 / 14A lighter documented profile than Modal
- Daytona8.0 / 14Fuller documented coverage on Observability & Auditability and Browser & Computer UseModal vs Daytona →Video
- E2B8.0 / 14Fuller documented coverage on Integrations & Tool Calling and Browser & Computer UseModal vs E2B →
- Unikraft6.0 / 14A lighter documented profile than Modal
- Composio9.5 / 14Adds documented Human Oversight & Guardrails
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded