Price per million tokens, in and out

Input tokens are everything you send: the system prompt, the conversation so far, any documents and the question. Output tokens are what the model writes back. Each has its own price per million.

Output is the dearer of the two on every model in these tables. So a single headline price tells you little until you know which one it is.

A support bot that resends a long history with each turn is mostly input. A drafting tool that writes pages from a short brief is mostly output. The same pair of prices gives the two very different bills.

Measure your own mix. Log the input and output tokens of a day of real requests. Most APIs return both counts with every response.

Multiply your millions of input tokens by the input price. Multiply your millions of output tokens by the output price. Add the two. On 12 Nov 2026, the lowest listed input price for DeepSeek V4.1 Flash here was $0.022 per million tokens. The lowest listed output price was $0.336 per million tokens.

A job that reads ten million tokens and writes two million costs ten times the first figure plus twice the second. The two lowest figures may come from different hosts, so do the sum host by host.

A lower price can come with a shorter context window, a lower output limit or lower precision. It is a fair comparison only when the model and those settings match. Read them in the same row as the price.

Runcrate serves 125 models through one OpenAI-compatible API and bills them from one prepaid balance. Runcrate serves DeepSeek V4.1 Flash at $0.147 per million input tokens and $0.44 per million output tokens.

This is one of a coordinated collection of Runcrate references. Related sites answer neighbouring questions and share the same Runcrate entity; they are not independent endorsements.