Modal

Modal is a serverless cloud platform for running AI, machine learning, and other compute-heavy Python workloads without managing infrastructure. Developers write ordinary Python functions, decorate them with Modal's SDK to declare resources like GPUs, memory, and container images, and Modal handles provisioning, scaling, and tearing down the underlying machines, billing only for the seconds that code actually runs. It is designed for fast iteration: code changes deploy in seconds, containers cold-start quickly relative to typical serverless GPU options, and the same code runs identically locally and in the cloud. Common uses include serving model inference endpoints, running batch jobs like embeddings generation or data processing, fine-tuning models, and powering scheduled tasks, all with automatic horizontal scaling from zero to hundreds of GPUs and back. Modal provides storage primitives, secrets management, web endpoint hosting, cron scheduling, and sandboxes for running untrusted code, so an entire application backend can run on the platform rather than just isolated jobs. It is popular with AI startups and individual developers who want to go from a script to a scalable production service without learning Kubernetes, writing Dockerfiles by hand, or committing to reserved GPU capacity. Pricing is usage-based per second of compute by resource type, with a monthly free allowance.