GPUs that appear when your function runs.
Modal runs Python functions on cloud GPUs with per-second billing and no infrastructure to manage — you decorate a function\, it executes on an H100\, and you pay only for the seconds it ran.
For bursty AI workloads that model is dramatically cheaper than reserving instances that idle. Rates run from $0.000164/sec on a T4 to $0.001972/sec on a B300\, and the Starter plan includes $30/month of free compute — enough for real experimentation.
Quick Information
Platform
Web
Pricing
Free credits + usage-based (from $0.000164/sec)
API
Available
Category
Coding & Development
Pros and Cons
Pros
- Per-second billing with no idle cost
- $30/month free compute on Starter
- No infrastructure to manage
- Scales to thousands of containers
- Academic and startup credits
Cons
- Python-centric
- Costs climb on sustained workloads
- Team plan has a $250 platform fee
- Cold starts on infrequent functions