Modal

Modal, built by Modal Labs, is a cloud platform for running AI workloads in production, including inference, training, batch processing, and sandboxed environments, with a Python-based SDK that lets engineers deploy to the cloud while staying in code. It offers sub-10ms latency LLM and multi-modal model inference, fine-tuning and multi-node training scaling up to 128 B200 GPUs, and isolated, ephemeral sandboxes for running agents and untrusted code. Modal auto-scales from zero to more than 1,000 GPUs instantly across multiple clouds, and includes built-in logging and monitoring for production readiness with globally distributed infrastructure and automated fleet health management. New users get $30 per month in free compute, with the rest priced on a usage-based, pay-per-second GPU model with no reserved capacity commitments.