Tuesday, October 27, 2026 · Edition updated 6:00 AM ETTUE, OCT 27 editionNext edition in Next in --:--:--
Intelligence Verdict
AI models · Mistral AI

Mistral Large 4

“Mistral Large 4 is a state-of-the-art, open-weight, general-purpose multimodal model with a granular Mixture-of-Experts” Mistral AI, in its own words

  • $0.68 / $2.09per million tokens in / out, the maker's price
  • 1Mtokens it reads at once
  • 0of 21 independent tests that place it near the top
The facts

Mistral Large 4 at a glance.

From Mistral AI's own pages, read Oct 20, 2026, so company-reported. A token is a word or part of a word. Prices can change; check the maker's page before you build on them.

Made by
Mistral AIBased in France
What it is for
“Mistral Large 4 is a state-of-the-art, open-weight, general-purpose multimodal model with a granular Mixture-of-Experts”Mistral docs: Mistral Large 4 model card, Oct 20, 2026 · in Mistral AI's words
Reads at once
1M tokensIts context window: how much text it can take in one go
Takes in
text, image
Download
No, online only
Size
1.05T parameters, 52B activeMistral docs: Mistral Large 4 model card, Oct 20, 2026
Price for developers

$0.68per million tokens in$2.09per million tokens out

Sale price

Sale price shown on the model card; regular price $1.36 in / $4.18 out. Cached input $0.07 (regular $0.14). Launched as a 'Public Preview'.

Mistral docs: Mistral Large 4 model card, Oct 20, 2026 · company-reported

Independent tests

Where it ranks.

Its position on each of the 21 tests the models page uses, by testers other than the maker. Epoch AI's and Vectara's scores are shown under their open licenses; for LMArena, Scale and Vals we give the position and link to their own leaderboard. Each list is a snapshot with its date.

All-round
TestRankScore
Epoch Capabilities IndexEpoch AI, Oct 20, 2026Not in the top 30
Vals IndexVals AI, Oct 19, 2026Not in the top 20
Coding
TestRankScore
LMArena WebDevLMArena, Oct 19, 2026Not in the top 20
SWE-Bench ProScale AI, Sep 22, 2026Not in the top 16
Vibe Code BenchVals AI, Oct 18, 2026Not in the top 20
LMArena coding chatsLMArena, Oct 15, 2026Not in the top 20
Writing and conversation
TestRankScore
LMArena textLMArena, Oct 15, 2026Not in the top 20
LMArena creative writingLMArena, Oct 15, 2026Not in the top 20
MultiChallengeScale AI, Feb 3, 2026Not in the top 15
Agents and computer use
TestRankScore
LMArena Agent ArenaLMArena, Oct 19, 2026Not in the top 20
Remote Labor IndexScale AINot in the top 16
Terminal-Bench 4.0Vals AI, Oct 19, 2026Not in the top 20
CUA-bench (games)Vals AI, Oct 13, 2026Not in the top 9
Math and science
TestRankScore
LMArena mathLMArena, Oct 15, 2026Not in the top 20
Humanity's Last ExamScale AINot in the top 17
ProofBenchVals AI, Oct 12, 2026Not in the top 20
Images and documents
TestRankScore
LMArena visionLMArena, Oct 19, 2026Not in the top 20
LMArena documentsLMArena, Sep 13, 2026Not in the top 20
More exam questions
TestRankScore
GPQA DiamondVals AI, Sep 1, 2026Not in the top 20
MMMU ProVals AI, Sep 1, 2026Not in the top 20
Sticking to the facts
TestRankScore
Vectara hallucination testVectara, Oct 16, 2026Not in our copy of the list

“Not in the top” means it isn't among the models in our copy of that list, which runs to the position shown; it may not have been tested. Vectara's test ranks by how rarely a model adds facts, so lower is better. Positions can be shared when scores tie. Epoch's range is its margin of uncertainty. Epoch AI and Vectara data: CC BY 4.0 and Apache-2.0.

For people, not developers

What Vibe costs.

Mistral AI's chat app, priced by the month in US dollars on its own pages. Which plans include Mistral Large 4 is on Mistral AI's pages.

Every chat app's plans
Vibe (formerly Le Chat)
PlanA month
FreeFree
Pro$14.99
Team$24.99 a seat
EnterpriseNot shown

Mistral AI: Pricing page, Oct 20, 2026 · company-reported

Our note: Announced as open-weight, but on 2026-10-20 weights were not yet downloadable: model card marks them 'Coming soon' and the announcement says 'We will release the weights by the end of the month.' No license named yet. The announcement describes it as 'a 1 trillion-parameter natively multimodal model with 52 billion active parameters'; the model card says 1.05T total / 52B active plus a 1.6B vision encoder. Quote is the opening of a longer sentence.