What will it cost?

How the sum is done

Every request is billed in three parts: new prompt tokens at the input price, repeated prompt tokens at the cached price, and the reply at the output price. A message costs the three added together.

What counts as a token

A token is a piece of a word. In English, a thousand tokens is about 750 words. A short chat message is 20 to 50 tokens. A page of text is around 500.

Why the repeat share matters so much

A chat sends the whole conversation again with every new message, and a coding agent resends the same files on every step. That repeated part is billed at the cached price, which is a fraction of the input price. In a long chat or an agent session, well over half the prompt is usually a repeat. Set the share to zero to see the price of a one-off request.

Keeping the bill down

• Pick the smallest model that does the job. The gap between the dearest and cheapest model here is more than twenty times on output.
• Cap the reply length with max_tokens. Output is the expensive side on every model.
• Put a spend cap on each key in the dashboard, so a loop cannot run past a figure you chose.

Keep reading

• Kimi K3.1 Pro, uncensored→
• GLM 5.3, uncensored→
• DeepSeek V4.1 Flash, uncensored→
• Qwen3.8 27B, uncensored→
• The uncensored AI API→
• Set up SillyTavern→
• Set up opencode→
• Set up Claude Code→
• Abliterated vs uncensored→
• Uncensored AI roleplay→

Updated 12 November 2026. Prices are read from the live price list.

Start chattingAll models and prices