What it is
Modal runs your Python code in containers in its cloud and scales them up and down with demand. You define functions and their hardware in code, from T4 and L4 GPUs to H100, H200, B200 and B300, and Modal handles scheduling, autoscaling and storage volumes. It is used for model inference, fine-tuning and evaluation, batch jobs, and sandboxes for coding agents.
It suits engineers who want GPUs without running Kubernetes. Billing is per second of compute: an H100 costs about $0.0011 a second. The Starter plan is free with $30 of compute a month; Team costs $250 a month plus compute. Containers stay warm for 60 seconds after the last request by default, and that idle time is billed.
Best for
- Deploying a custom or open model behind an API without managing servers
- Bursty GPU workloads that should scale to zero
- Running untrusted or agent-written code in isolated sandboxes