Pay for what runs.
One balance of credits pays for open models and for sandboxes.
Inference
Each model has a posted price for each million tokens. You pay only for the tokens that you use.
- Start free
Free
$0
- The free tier, on spare capacity.
- A verified phone number.
- Daily limits on each model.
- Price of each model
Pay as you go
Per token
- The posted price of each model, by region and service tier.
- Prepaid credits, paid by card.
- More credits raise your rate limits.
- Talk to an engineer
Committed spend
Volume rates
- An order form and a monthly invoice, net 30.
- Dedicated endpoints: reserved GPU capacity for one model.
Sandboxes
Cloud computers on demand. A plan sets how many run at once, and its price comes back as credits for the month.
- Choose Starter
Starter
$15 a month
100 sandboxes at once
- Choose Pro
Pro
$95 a month
300 sandboxes at once
- Choose Scale
Scale
$495 a month
1,000 sandboxes at once
- Choose Fleet
Fleet
$1,995 a month
2,000 sandboxes at once
Enterprise
More than 2,000 sandboxes at once, on dedicated hosts, in your own cluster or on your own servers, at volume prices.
Talk to an engineerStart with a 7-day trial: add a card, and get $0.90 of sandbox time with 2 sandboxes at once.
Every plan pays the same rates, $0.018 an hour for 2 vCPUs and 4 GiB, for each second that a sandbox runs. Usage past the credits of a plan takes your prepaid credits. Sizes and prices