Provider pricing

Modal AI API pricing

Modal is a serverless platform for running Python on GPUs with per-second billing and sub-second cold starts. Teams use it to serve custom fine-tunes of Llama, Qwen, and DeepSeek, run distributed training, and orchestrate batch inference — all defined as plain Python functions with no Dockerfile or YAML required.

Founded2021
HeadquartersNew York, NY
Tracked modelsDirectory

Best for

Engineering teams serving custom models or running batch inference where flexibility beats a fixed catalog.

Strengths

  • Per-second GPU billing with no idle charges
  • Sub-second cold starts via memory snapshotting
  • Define infra in Python — no Dockerfiles, no YAML
  • Built-in scheduling, queues, and cron for batch workflows