Modal runs CPU and GPU containers for functions, batch jobs, and model serving. It suits your own inference code when compute, storage, and transfer costs need separate accounting.
As of 2026-10, Starter has a $0 base fee, $30 in monthly compute credits, and up to three seats. H100 SXM5 costs $0.001097 per second, CPU $0.0000131 per core-second, and memory $0.00000222 per GiB-second. One TiB per month of egress and one TiB-month of volume are included; extra egress is $0.04 per GiB.
The GPU rate alone omits CPU, memory, volume, and egress. The demo applies monthly credits to H100 GPU time only as a simplified example.
When to use
Choose it when you need to run your own model code and GPU jobs. Compare Replicate if calling a published model is enough.