Mistral Large 4 · Mistral

What does Mistral Large 4 cost to run in a coding agent?

Mistral Large 4 costs about $0.042 per average coding-agent request on our measured traffic, at list prices of $1.36 per million input and $4.18 per million output tokens. It scores 38 on the Artificial Analysis Intelligence Index, 20th of the 48 models we track. Source: Artificial Analysis (artificialanalysis.ai)

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 Mistral Large 4 in your agent from $19/month

See plans

What Mistral Large 4 costs per coding-agent request

Cost at list price	Mistral Large 4
Prompt under 100K tokens50% of requests, 87% of input cached	$0.020
Prompt of 100K to 272K39% of requests, 88% of input cached	$0.051
Prompt over 272K tokens11% of requests, 86% of input cached	$0.12
Average request	$0.042
100 requests	$4.24
Median session16 requests; a session ends after 10 idle minutes	$0.55
A month of regular usemedian of 20 users active on 5+ days	$64.65

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.

Mistral Large 4 benchmarks, speed and latency

Artificial Analysis	Value	Rank
Intelligence Index10 evaluations, 0 to 100	38	20 of 48
Terminal-Bench 4.0agentic coding in a terminal	26.8%	15 of 39

Source: Artificial Analysis (artificialanalysis.ai). Measured variant: Mistral Large 4 Preview, speed on Mistral. Fetched October 13, 2026. Rank is among the models on this site, older versions included. Artificial Analysis runs Terminal-Bench 4.0 in its own setup, so its scores differ from the tbench.ai leaderboard.

About Mistral Large 4

Mistral Large 4 is Mistral's open-weight flagship, released October 6, 2026 as a public preview. Mistral describes a mixture-of-experts model with 52B active and 1.05T total parameters plus a 1.6B vision encoder, built for reasoning, coding and agent work, with a 1M-token context. It is a large step up from Large 3: Artificial Analysis scores the preview 38 on its Intelligence Index and 26.8% in its Terminal-Bench 4.0 run, against 9 and 0% for Large 3. That still sits below the frontier models, and at list price it costs $0.042 per request in our measurements, about three and a half times Large 3. Mistral currently sells it at 50% off, with no end date given.

Sources: Mistral model page · Artificial Analysis

Mistral Large 4 FAQ

How much does Mistral Large 4 cost per coding-agent request?

$0.042 on average at list price, $4.24 per 100 requests, and $0.55 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 Mistral Large 4 charge more for long prompts?

No. Mistral Large 4 bills the same rate per token up to its 1M-token context window.

Is Mistral Large 4 good for AI agents?

Open-weight, EU-provenance agent work that needs long context and image input. On our scoring, which is calibrated to public benchmarks, it rates 7.1/10 for agentic ability and 7.6/10 for output quality. Open weights with a 1M-token context and image input.

What is Mistral Large 4's context window?

Mistral Large 4 has a 1.0M-token context window, large enough to hold sizeable codebases or document sets in a single run. Mistral is the maker.

Is Mistral Large 4 open source?

Yes. Mistral Large 4 ships with open weights, so you can self-host it or run it through a hosted provider such as OpenRouter. Check the license before commercial use.

What are the downsides of Mistral Large 4?

The main trade-offs: Still a public preview, so behavior and price can change; 26.8% in Artificial Analysis's Terminal-Bench 4.0 run, well below the frontier models; The 50% sale price has no published end date. It's strongest for open-weight, EU-provenance agent work that needs long context and image input.