What is Modal?
A serverless GPU cloud platform. Python decorators define cloud functions, with automatic GPU allocation, scaling, and billing.
TL;DR: Serverless GPU cloud. Python decorator → cloud A100/H100. Auto-scaling, per-second billing, no Docker/K8s. Simplest path for inference / fine-tuning / batch.
Best for: ML engineers who need cloud GPUs but don't want to manage infrastructure.
Core Features
1. GPU Selection — T4/A10G/A100/H100, multi-GPU supported
2. No Dockerfile — container defined in Python code
3. Web Endpoints — deploy an API in one line
4. Per-Second Billing — no minimum spend
FAQ
Q: Cold start? A: ~30s first time; warm containers < 1s.