Modal

Serverless platform for AI workloads. Long-running functions, GPU support, and scale-to-zero. Built for agents that need to think for minutes.