Modal

Modal is a serverless compute platform for running Python code on cloud CPUs and GPUs with per-second billing, autoscaling, and no infrastructure management. It is popular for AI inference, model training, batch processing, and data pipelines, appealing to developers who want to define infrastructure as Python code rather than manage clusters and orchestration.

Modal is a serverless cloud built for compute-heavy Python workloads. You define functions in Python, decorate them to run on Modal, and the platform provisions containers on CPUs or GPUs on demand, scaling from zero to many instances and billing by the second of actual usage. It removes the need to manage clusters, Dockerfiles-by-hand, or orchestration, letting developers move quickly from local code to scalable cloud execution. Modal is widely used for AI and ML: running inference endpoints, fine-tuning and training models, batch processing, embeddings generation, and data pipelines. It provides fast container start times, GPU access (including high-end accelerators), scheduled jobs, web endpoints, storage, and secrets management, all defined in code. Its programmable, developer-first model appeals to teams that want infrastructure expressed as Python rather than YAML. The platform suits engineers who are comfortable writing code and want flexible, general-purpose serverless compute rather than a fixed model API. It offers monthly free compute credits to start, then usage-based pricing tied to CPU, GPU, and memory time. For teams whose needs go beyond calling a hosted model, Modal is a powerful, flexible foundation.