Mistral Large 4
Mistral Large 4 is capable in long context; and behind the leaders in coding, multimodal tasks, reasoning, agentic tasks, factuality, and math. Too few results yet to rate safety, multilingual tasks, or instruction following.
Capability profile
Bars show the model's median result as a share of the leading model's, per capability. Select a row to see its results.
Too few results yet to rate Safety, Multilingual or Instruction Following.
Price
$0.70input$2.10outputper million tokens
From Artificial Analysis · 2 providers tracked · All prices
Evidence
63results on47benchmarks
- 15 independently verified
- 21 aggregator
- 12 vendor-reported
- 15 cross-referenced
From 12 sources · latest Nov 20, 2026 · How verification works
Mistral Large 4 benchmark results
63 results on 47 benchmarks, grouped by capability. Each bar is the result as a share of the capability leader's; every result links to its source.
17.5% behind the leader1 of 3 ranked benchmarks measured
- 81.33Oct 7, 2026
26.1% behind the leader5 of 10 ranked benchmarks measured
- 90.20Nov 5, 2026
- 1569.76Oct 9, 2026
- LiveBench · Coding77.16Oct 18, 2026livebench_coding@2026-06-25
- SciCode54.17Oct 7, 2026aa_scicode
- LiveBench · Agentic Coding57.17Oct 18, 2026livebench_agentic_coding@2026-06-25
Show 1 more coding resultHide 1 coding result
- 91.80Nov 19, 2026
27.3% behind the leader1 of 6 ranked benchmarks measured
- MMMU-Pro76.42Oct 7, 2026aa_mmmu_pro
Show 1 more multimodal resultHide 1 multimodal result
- 1244.00Oct 20, 2026
28.0% behind the leader3 of 6 ranked benchmarks measured
- LiveBench · Reasoning83.94Oct 18, 2026livebench_reasoning@2026-06-25
- Humanity's Last Exam35.03Oct 7, 2026aa_hle
- 10.00Nov 20, 2026
Show 1 more reasoning resultHide 1 reasoning result
- 11.14Nov 20, 2026
29.0% behind the leader3 of 7 ranked benchmarks measured
- 84.50Nov 19, 2026
- 46.20Oct 7, 2026
- Terminal-Bench 4.024.75Nov 20, 2026
Show 1 more agentic resultHide 1 agentic result
- Terminal-Bench 4.023.23Nov 20, 2026
36.2% behind the leader3 of 4 ranked benchmarks measured
- AA-Omniscience · Non-hallucination58.05Oct 7, 2026omniscienceNonHallucination
- AA-Omniscience · Accuracy25.82Oct 7, 2026omniscienceAccuracy
- 22.40Oct 10, 2026
43.2% behind the leader5 of 5 ranked benchmarks measured
- LiveBench · Mathematics93.58Oct 18, 2026livebench_math@2026-06-25
- 58.60Oct 16, 2026
- 26.83Oct 16, 2026
0 of 3 ranked benchmarks measured
- LiveBench · Instruction Following64.82Oct 18, 2026livebench_instruction_following@2026-06-25
More results
Benchmarks outside the capability baskets. They are not ranked against a leader.
- AA Intelligence23.00Nov 14, 2026Artificial Analysis Intelligence Index
- AA Intelligence26.00Oct 30, 2026Artificial Analysis Intelligence Index
- livebench_data_analysis76.54Oct 18, 2026livebench_data_analysis@2026-06-25
- livebench_language49.64Oct 18, 2026livebench_language@2026-06-25
- AA Intelligence25.00Oct 6, 2026Artificial Analysis Intelligence Index
- Terminal-Bench-Science 0.14.29Nov 20, 2026
Show 33 more resultsHide 33 results
- AA Intelligence38.38Oct 7, 2026aa_intelligence_index
- -5.30Oct 7, 2026
- AgentHarm (cyber subset): 100 minus harm score100.00Nov 5, 2026
- Beyond-AIME (internal)90.60Nov 5, 2026
- 90.60Nov 6, 2026
- BrowseComp subset85.00Nov 6, 2026
- BrowseComp subset (internal)85.00Nov 5, 2026
- 63.10Nov 19, 2026
- 49.80Oct 6, 2026
- CritPt v2 (Hard Subset)26.00Nov 5, 2026
- CritPt v2 Hard27.00Nov 20, 2026
- CritPt v2 Hard22.50Nov 20, 2026
- 82.00Nov 19, 2026
- CyberGym-E2E-AA79.00Nov 5, 2026
- 61.70Nov 19, 2026
- 65.20Nov 5, 2026
- 54.70Nov 19, 2026
- GDP (Surge AI)17.60Nov 20, 2026gdp-pdf
- GDP (Surge AI)16.80Nov 20, 2026gdp-pdf
- GDP (Surge AI)18.60Nov 19, 2026GDP.pdf
- GDP (Surge AI)17.60Nov 5, 2026GDP.pdf
- HLE-Diamond17.30Nov 20, 2026
- HLE-Diamond13.90Nov 20, 2026
- HLE-Diamond17.30Nov 5, 2026
- Human Evaluation (coding quality, Surge AI)3.74Oct 16, 2026
- Human evaluation coding quality3.74Nov 17, 2026
- JailbreakBench (cyber subset): safe-response rate97.30Nov 5, 2026
- StrongREJECT (cyber subset): refusal rate92.60Nov 5, 2026
- Terminal-Bench 4.041.00Nov 19, 2026
- Terminal-Bench 4.028.30Oct 6, 2026Terminal-Bench 4
- Terminal-Bench 4.041.00Nov 5, 2026
- Terminal-Bench 4.0 (AA Coding Agent Index)41.00Nov 5, 2026
- Terminal-Bench-Science 0.13.33Nov 20, 2026
Mistral Large 4: common questions
Who makes Mistral Large 4?
Mistral Large 4 is made by Mistral.
When was Mistral Large 4 released?
Mistral Large 4 was released on Oct 6, 2026, according to Mistral's own announcement.
What is Mistral Large 4 good at?
Mistral Large 4 is capable in long context; and behind the leaders in coding, multimodal tasks, reasoning, agentic tasks, factuality, and math. Too few results yet to rate safety, multilingual tasks, or instruction following.
How much does Mistral Large 4 cost?
Mistral Large 4 costs $0.70 per million input tokens and $2.10 per million output tokens, according to Artificial Analysis. We track its price at 2 providers. At a mix of three input tokens to one output token, it costs more than 58% of the 337 priced models we track.
How many benchmarks has Mistral Large 4 been tested on?
We track 63 results for Mistral Large 4 on 47 benchmarks from 12 sources, 15 of them independently verified. The latest was recorded on Nov 20, 2026.
About this record
Where Mistral Large 4's numbers come from, and every name it appears under.
- Tracked since
- Oct 6, 2026
- Newest source mention
- Nov 6, 2026
Where the results come from
Verification: 63 scores · 15 independently verified · 21 aggregator-attributed · 15 vendor cross-reference · 12 vendor-reported. How these tiers are assigned
From 12 sources on 7 sites. Artificial Analysis supplies 24 of them; the 15 independently verified results come from 4 sites. Bars are coloured by trust tier.
- artificialanalysis.ai24
- alphaxiv.org15
- api.llm-stats.com8
- livebench.ai7
- mistral.ai4
- epoch.ai3
- datasets-server.huggingface.co2
Also known as
How our sources name Mistral Large 4 at each reasoning setting.
| Setting | Short form | API id |
|---|---|---|
| medium | mistral large 4 (medium) | mistral-large-4-medium |
| high | mistral large 4 (high) mistral large 4 preview (high) | — |