One subscription for every model. Credits regenerate over the week and apply at discounted model rates.See how rate limits work →
Turn on fast mode and your requests run up to 2× faster. While it’s on, each request uses 2× as much of your rate limit — great for interactive sessions where you’re waiting on every token.
Built for background agents: infinite mode requests never count against your rate limit, in exchange for running much slower. Kick off an agent and let it grind away in the background. (Currently DeepSeek V4.1 Flash only.)
Fast mode and Infinite mode are included with the Big plan
Your plan's weekly credit quota uses tokens at discounted rates. You can always buy more tokens at usage-based pricing rates, or subscribe to a higher-tier plan to increase your token limits.See how many tokens are in a prompt →
| Model | log scale | Tokens/mo | API $/mo |
|---|---|---|---|
| 7.32B | $204 | ||
| 7.32B | $234 | ||
| 2.70B | $251 | ||
| 1.02B | $55 | ||
| 968M | $88 | ||
| 391M | $56 | ||
| 248M | $100 | ||
| 71.9M | $59 |