Tokens in. Cents out.
No subscription, no seats, no annual contract. Every request is metered from the model's token counts and billed against your balance.
| Model | Input / 1M | Cached / 1M | Output / 1M |
|---|---|---|---|
| Unrestrict DeepSeek 1017 | $0.2143 | $0.0143 | $0.8571 |
| Unrestrict Gemma 4 26B A4B Heretic | $0.18 | $0.18 | $0.57 |
| Unrestrict Qwen 3.8 27B Derestricted | $0.15 | $0.15 | $0.90 |
| Unrestrict GLM 5.3 Flash Cybersecurity | $0.22 | $0.22 | $0.75 |
| Unrestrict Apex Flash 1 Abliterated | $0.22 | $0.22 | $0.75 |
Estimate a request
The whole bill, in one line.
Input tokens, output tokens, and nothing else. No platform fee, no per-request charge, no charge for cancelled streams.
Questions people actually ask
If yours is missing, the docs are shorter than this page.
How is usage counted?
From the token counts the model returns. Input and output are priced separately. A failed request is not billed; a cancelled stream is billed for the tokens already produced.
Do keys expire?
No. Keys stay active until you revoke them. Create one per project and revoke it the moment it leaks.
Is there a rate limit?
Requests are limited by concurrency, not by a daily quota. Your balance is the only spending limit.
Can I bring my own prompt?
Yes. There is no hidden system prompt. The messages you send are the messages the model sees.
What happens when the balance runs out?
Requests return a 402 until you top up. Nothing is deleted and no subscription renews behind your back.