What is Modal?
Serverless GPUs that autoscale for training and inference.
Who is Modal best for?
Modal is best for Python developers who run inference and fine-tuning on cloud GPUs without managing servers.
Model serving with spiky traffic
Inference endpoints scale down to zero between requests and burst with demand, so quiet hours do not pay for idle GPUs.
Fine-tuning and training runs
Jobs run on anything from one GPU to multi-node clusters of up to 128 B200s, with the hardware set in the same Python code as the job.
Large batch jobs
Evals, embeddings and dataset generation fan out across thousands of GPUs in parallel without a separate job scheduler to run.
Sandboxes for AI agents
Isolated environments start on demand for coding agents and reinforcement learning rollouts, which suits products that run untrusted code.
How much does Modal cost?
Modal's Starter plan is free with $30 of compute each month, Team costs $250 a month plus usage with $100 of compute included, and GPUs are billed by the second, from $0.59 an hour for an Nvidia T4 to $7.10 an hour for a B300.
- $30 of free compute a month
- 3 workspace seats
- Up to 100 containers and 10 GPUs at once
- $100 of free compute a month
- Unlimited seats
- Up to 5,000 containers and 50 GPUs at once
Compute is billed by the second: an Nvidia H100 costs $3.95 an hour, an A100 80 GB $2.50, an L4 $0.80 and a T4 $0.59; a CPU core costs $0.0473 an hour and memory $0.008 per GiB an hour. Volumes cost $0.09 per GiB a month after 1 TiB free. Choosing a region costs 1.15 to 1.75 times the base rate and non-preemptible runs 3 times; sandboxes and notebooks have their own, higher CPU and memory rates. Enterprise is priced on request, with volume discounts, SSO, audit logs and HIPAA. Startups can apply for free credits and academics for up to $10,000. Full pricing on modal.com. We checked these prices on .
How do readers rate Modal?
No one has rated Modal yet. Be the first.
Sign in to rate. One rating per account.
What are the best Modal alternatives?
Two tools on Recaply do Modal's job.
Questions about Modal
What is Modal?
Serverless GPUs that autoscale for training and inference.
Who is Modal for?
Modal is best for Python developers who run inference and fine-tuning on cloud GPUs without managing servers.
Is Modal free?
Yes, within limits. Modal's Starter plan costs $0 a month and includes $30 of compute every month with no end date, which covers about 7.5 hours on an Nvidia H100 or about 50 hours on a T4. Anything beyond that is billed at the per-second rates.
How much does Modal cost?
Modal's paid plans are Team at $250 per month plus compute.
What are the best Modal alternatives?
The closest alternatives on Recaply are Tensorfuse and Replicate.
Where can I try Modal?
On its official website, modal.com.
How does Modal bill GPU time?
Modal bills GPUs, CPU cores and memory by the second while your code runs, and nothing while it sits idle. An Nvidia H100 costs $3.95 an hour and an A100 80 GB $2.50 an hour; pinning a specific region raises the rate by 15% to 75%.







