Codestral · Mistral

What does Codestral cost to run in a coding agent?

Codestral costs about $0.0070 per average coding-agent request on our measured traffic, at list prices of $0.30 per million input and $0.90 per million output tokens. Mistral built it for autocomplete and quick edits, and no independent agent benchmark lists it.

List prices: Mistral API pricing, checked 2026-10-13. Request costs: Standard Compute gateway data, 79,301 coding-agent requests from 75 users, Sep 13 to Oct 12, 2026. How we compute this

Run current Claude, GPT, Gemini and open models in your agent

Or let the router send each step to the cheapest model that can do it, on providers we vet for price, uptime and speed. Plans from $19 a month.

See plans

What Codestral costs per coding-agent request

Cost at list price	Codestral
Prompt under 100K tokens50% of requests, 87% of input cached	$0.0043
Prompt of 100K to 272K39% of requests, 88% of input cached	$0.011
Prompt over 272K tokens11% of requests, 86% of input cached	n/a
Average request	$0.0070
100 requests	$0.70
Median session16 requests; a session ends after 10 idle minutes	$0.084
A month of regular usemedian of 20 users active on 5+ days	$8.91

Each logged request priced at the vendor’s list rates for its own token counts, including cache reads and any higher rate for long prompts. Average output: 698 tokens per request. 13% of requests were larger than Codestral's 256K context window and are left out of its figures.

What to watch for

Its 256K window is smaller than 13% of our coding-agent requests, which are left out of the figures above, so long agent sessions will hit the limit. For agent loops, Devstral 2 and current general models are the closer fit.

About Codestral

Codestral (2508) is Mistral's low-latency code model with a 256K context, aimed at autocomplete and quick edits rather than agentic planning. No independent agent benchmark lists it. It costs $0.0070 per request in our measurements, but its 256K context is smaller than 12.5% of the coding-agent requests we measured.

Codestral FAQ

How much does Codestral cost per coding-agent request?

$0.0070 on average at list price, $0.70 per 100 requests, and $0.084 for a median 16-request session. That is 79,301 real coding-agent requests (Sep 13 to Oct 12, 2026) priced at Mistral's list rates; the median request sent 99,534 input tokens, 87% of input was read from cache, and requests returned 698 output tokens on average.

Does Codestral charge more for long prompts?

No. Codestral bills the same rate per token up to its 256K-token context window, which is smaller than 13% of the coding-agent requests we measured.

Is Codestral good for AI agents?

Fast completions and lightweight code edits. On our scoring, which is calibrated to public benchmarks, it rates 5.0/10 for agentic ability and 3.4/10 for output quality. Built for low-latency code completion.

What is Codestral's context window?

Codestral has a 256K-token context window. Mistral is the maker.

Can I self-host Codestral?

No. Codestral is a proprietary, hosted-only model from Mistral. You access it through an API, such as OpenRouter, rather than running the weights yourself.

What are the downsides of Codestral?

The main trade-offs: Not built for agentic planning; No independent agent benchmark results. It's strongest for fast completions and lightweight code edits.