Modal

What it is for

Modal is a signed-out serverless GPU/CPU playground plus a Python runtime: the live still (see the cover: Run your first function, import modal, @app.function, def square(x=2), green Run, terminal modal run get_started.py) is the getting-started canvas. Use it when the job is run a containerized function on burst GPUs, not when you want a chat window.

It is not open source as a product (the Python client is on GitHub). Billing is Starter $0 + compute or Team $250 + compute.

What makes it unique: the cover is Modal’s own Playground chrome — editor + Run — not a GPU collage.

Key features

• Playground editor — modal.App("example-get-started"), @app.function, Run (cover).
• Pay per second — GPUs, CPU cores, and memory. No idle reservation on the public table.
• Starter included compute — $30 / month of compute on the $0 Starter plan.
• Team — $250 / month + compute; $100 included compute; unlimited seats.
• Sandboxes / Notebooks — billed at non-preemptible rates (3× CPU/memory list) on the same page.

Who it's for

Best fit: Python teams with spiky GPU/CPU jobs who will read a per-second table.

Poor fit: people who want a hosted chat playground (Together AI) or a one-click image studio (Replicate).

Pricing

Verified 2026-10-21 (Asia/Shanghai) from modal.com/pricing. Re-check before you buy.

• Starter — $0 / month + compute. Up to 3 seats. Included compute $30 / month. 1-day logs. 5 deployed crons.
• Team — $250 / month + compute. Unlimited seats. Included compute $100 / month. 30-day logs. RBAC. Custom domains.
• Enterprise — custom compute. HIPAA / SSO / audit logs on that column.
• Example GPU list (per second) — H100 SXM5 $0.001097 / sec; A100 80 GB $0.000694 / sec; T4 $0.000164 / sec.
• CPU — $0.0000131 / core / sec. Memory — $0.00000222 / GiB / sec. Volumes — $0.09 / GiB / mo (1 TiB/mo free).

Shared Endpoints are per token and are not covered by the Starter $30 compute grant (FAQ, effective 2026-09-01).

Limitations / costs

• $0 is the seat, not the GPU. H100 still bills by the second after the $30 grant.
• Starter is 3 seats. A fourth teammate needs Team.
• Token endpoints are extra. They do not come out of included compute.
• Not a chat product. The Playground runs Python, not a model picker for consumers.

Alternatives

• Replicate — per-prediction model playground (e.g. Flux Schnell $0.025 / image).
• Together AI — hosted inference with public token cards.
• vLLM — Apache-2.0 serving engine you run yourself.

Official tutorial

Official Modal channel video (linked from modal.com): Getting started with Modal (oEmbed author: Modal).

Written Introduction plus the signed-out Playground.

Website

https://modal.com

Repository

https://github.com/modal-labs/modal-client

Source

Catalog row curated for AI Tools Nav. Cover is the live signed-out Modal Playground (editor + Run), not a marketing GPU wall. Dollars checked on modal.com/pricing (Starter $0 + $30 compute; Team $250; H100 $0.001097/sec). Tutorial: Getting started with Modal on the official Modal channel (oEmbed).

Last verified

2026-10-21 (Asia/Shanghai) editorial check. Starter is $0 + $30 included compute. Shared Endpoints are token-billed from the first request.