No subscriptions, no seat fees, no minimums. Buy credits once and spend them across any model — priced transparently per million tokens.
Prices are in credits per 1M tokens. Cache read and cache write apply to models that support prompt caching.
| Model | Context | Input | Output | Cache read | Cache write |
|---|---|---|---|---|---|
| DeepSeek V4.1 Flash | 1M | 0.0030000 | 0.0127000 | 0.0000800 | 0.0000000 |
Prices shown in credits per 1M tokens. Final cost per request is computed from the exact tokens reported in the response.
The same rates, laid out per model for easy comparison.
Context window: 1M tokens · credits per 1M tokens
Credits are a prepaid balance denominated in tokens. Each request reserves, measures and settles — so you never overspend.
When a request arrives, we estimate the maximum cost and reserve it from your balance before serving it.
After the response completes we read the real token usage — input, output, cache read and cache write — from the response.
We charge the exact cost for those tokens and release the unused reservation back to your balance. Failed requests are refunded in full.
Add credits by submitting a top-up request from your dashboard. Requests are reviewed and approved by an operator — there is no automatic card charge.
You are never billed for tokens you did not receive.