Modal runs Python functions on serverless CPUs/GPUs with per-second billing — custom model inference, batch jobs and ML pipelines without managing a single machine.
Modal
Serverless GPUs: run any Python/ML workload with @app.function
modal.com · $30/mo free credit; per-second GPU billing
Capabilities
Capabilities
- Serverless A100/H100s, per-second billing, cron, web endpoints; $30/mo free compute
Best for
- Custom-model features a hosted API doesn't offer (fine-tunes, niche pipelines)
Not for
- Mainstream models (first-party APIs and fal/Replicate are simpler)