Modal

Modal provides elastic GPU capacity with serverless compute for AI, ML, and data teams. Spin up GPU-enabled containers in as little as 1 second, autoscale to hundreds of GPUs and back to zero without managing infrastructure. Features pay-per-use pricing with no quotas or reservations. Custom infrastructure enables 2-4 second cold starts. Used for ML inference, fine-tuning, batch jobs, and GPU-accelerated tasks. Raised $80M in 2025. Most flexible serverless GPU provider with ergonomic Python SDK for arbitrary code execution.