MistralReleased Oct 6, 2026
Open weights

mistral-large-4

Strongest in Knowledge Work (#37 of 186) and Programmatic Reasoning (#45 of 187); weakest in Computer Use (#97 of 186). Abstract Reasoning is estimated, because no results are published yet.

Composite IQ ?
118
#62 of 186 ranked models · ahead of 67%
World (Other)'s top-ranked model
Effective cost
$4.69/1M
$0.70 in · $2.10 out
Speed
116tok/s
23s typical response
Context
1Mtokens
1,000,000 tokens in
Evidence
6of 7 dims
19 of 39 benchmarks

IQ profile

Seven dimensions, averaged equally into Composite IQ
Compare shape with
100 avg human130160● Abstract99#87 · est.● Math121#64● Academic128#63● Coding132#45● Computer Use110#97● Reliability114#78● Knowledge Work126#37
MeasuredEstimated95% range
#1 in dimensionTop-10 median

IQ dimensions

Ranks are among models scored in each dimension. Ranges are model-based estimates, not calibrated confidence intervals.

DimensionIQPosition · 70–160RankEvidencevs Top 10
99estimated
#87 / 226
0/5
-33.5
121measured
#64 / 186
4/5
-28.5
128measured
#63 / 228
4/6
-7
132measured
#45 / 187
2/6
-19
110measured
#97 / 186
1/5
-36.5
114measured
#78 / 188
4/6
-21.5
126measured
#37 / 186
4/6
-6
mistral-large-495% range (model-based, uncalibrated)Top-10 median#1 in dimension100 = average human

Benchmark results

19 of 39 Composite IQ benchmarks have a published result for mistral-large-4. Benchmarks without a result stay listed so gaps are visible.

Abstract Reasoning IQ 99 · estimatedNo published results0/5
No published results yet for Abstract Reasoning. The 99 shown is estimated from mistral-large-4's other dimensions and will be replaced as results come in. How estimates work
ARC-AGI-3ARC-AGI-2ARC-AGI-1Chess PuzzlesMystery Game Puzzles
Mathematical Reasoning IQ 121 · measuredBest: #34 of 70 on MathArena4/5
BenchmarkScorePercentileRankLeader · source
MathArena
Full source leaderboard reconciliation · Oct 28
26.4%
#34 / 70
Leader gpt-6.2-astra · 92.9%
Full source leaderboard reconciliation · Oct 28
FrontierMath Tier 1-3
Full source leaderboard reconciliation · Oct 23
58.6%
#46 / 90
Leader fable-5.2 · 95.1%
Full source leaderboard reconciliation · Oct 23
FrontierMath Tier 4
Full source leaderboard reconciliation · Oct 23
26.8%
#43 / 78
Leader fable-5.2 · 100%
Full source leaderboard reconciliation · Oct 23
ProofBench 1.1
Full source leaderboard reconciliation v1.1 · Oct 27
10.0%
#45 / 49
Leader fable-5.1 · 100%
Full source leaderboard reconciliation v1.1 · Oct 27
Not yet reported
AIME
Academic Reasoning IQ 128 · measuredBest: #40 of 167 on SciCode4/6
BenchmarkScorePercentileRankLeader · source
SciCode
Artificial Analysis v4.3.2 · Oct 27
54.2%
#40 / 167
Leader opus-5.5 · 66.9%
Artificial Analysis v4.3.2 · Oct 27
CritPt
Artificial Analysis v4.3.2 · Oct 27
10.6%
#60 / 159
Leader Muse Wonder Full · 32.9%
Artificial Analysis v4.3.2 · Oct 27
Humanity's Last Exam
Artificial Analysis v4.3.2 · Oct 27
35.0%
#69 / 167
Leader opus-5.5 · 61.4%
Artificial Analysis v4.3.2 · Oct 27
MMMU-Pro
Artificial Analysis v4.3.2 · Oct 27
76.4%
#45 / 90
Leader fable-5.1 · 90.6%
Artificial Analysis v4.3.2 · Oct 27
Not yet reported
GPQA DiamondMMLU-Pro
Programmatic Reasoning IQ 132 · measuredBest: #28 of 109 on Terminal-Bench 4.02/6
BenchmarkScorePercentileRankLeader · source
Terminal-Bench 4.0
Artificial Analysis v4.3.2 · Oct 27
26.8%
#28 / 109
Leader Kimi K3.1 Pro · 67.2%
Artificial Analysis v4.3.2 · Oct 27
IOI V2
Full source leaderboard reconciliation v2 · Oct 27
45.3%
#31 / 43
Leader Gemini 4 Argon · 100%
Full source leaderboard reconciliation v2 · Oct 27
Not yet reported
LiveCodeBenchProgramBenchSWE-rebenchMirrorCode
Computer Use IQ 110 · measuredBest: #59 of 77 on Arena.ai Agent Arena1/5
BenchmarkScorePercentileRankLeader · source
Arena.ai Agent Arena
Arena.ai Agent Arena · Nov 19
-7.1
#59 / 77
Leader fable-5.2 · 27.4
Arena.ai Agent Arena · Nov 19
Not yet reported
BrowseCompOSWorld-VerifiedMCP AtlasAgents' Last Exam
Reliability IQ 114 · measuredBest: #17 of 96 on BullshitBench v24/6
BenchmarkScorePercentileRankLeader · source
BullshitBench v2
Full source leaderboard reconciliation · Oct 23
68.0%
#17 / 96
Leader opus-4.8 · 94.0%
Full source leaderboard reconciliation · Oct 23
AA Long Context Reasoning v1.1
Artificial Analysis v4.3.2 · Oct 27
81.3%
#36 / 154
Leader kimi-k3 · 88.7%
Artificial Analysis v4.3.2 · Oct 27
AA Omniscience
Artificial Analysis v4.3.2 · Oct 27
-5.3
#81 / 158
Leader opus-5.5 · 46.4
Artificial Analysis v4.3.2 · Oct 27
SimpleQA Verified
Full source leaderboard reconciliation · Oct 23
20.0%
#79 / 86
Leader fable-5.2 · 94.0%
Full source leaderboard reconciliation · Oct 23
Not yet reported
IFBenchFACTS Grounding
Knowledge Work IQ 126 · measuredBest: #25 of 112 on AutomationBench-AA4/6
BenchmarkScorePercentileRankLeader · source
AutomationBench-AA
Artificial Analysis v4.3.2 · Oct 27
59.9%
#25 / 112
Leader Gemini 4 Carbon · 77.8%
Artificial Analysis v4.3.2 · Oct 27
GDP.pdf
Artificial Analysis v4.3.2 · Oct 27
18.6%
#34 / 111
Leader Muse Wonder Full · 32.6%
Artificial Analysis v4.3.2 · Oct 27
AA-Briefcase v1.1 Elo
Artificial Analysis v4.3.2 · Oct 27
1,391
#37 / 116
Leader sonnet-5.5 · 1,816
Artificial Analysis v4.3.2 · Oct 27
GDPval-AA v2.1 normalized score
Artificial Analysis v4.3.2 · Oct 23
46.1%
#40 / 124
Leader opus-5.5 · 68.3%
Artificial Analysis v4.3.2 · Oct 23
Not yet reported
AA-AnalystAgentEnterpriseOps-Gym-AA

Release & availability

Facts from the publisher's announcement and documentation.

Released
Developer
Mistral
Access
Mistral Large 4 is available through the Mistral API and as downloadable weights on Hugging Face. The improved weights were released October 29, 2026.
Weights
Open weightsDownloadable and self-hostable
Inputs
textimage
Context
1,000,000 tokens
API price
$0.70 in · $2.10 outper 1M tokens · full pricing

What Mistral says

  • General-purpose multimodal model supporting reasoning, coding and agentic workflows.
  • Improved Large 4 weights released October 29 with low, medium and high reasoning modes.

Limitations

  • The Mistral Ecosystem Open-Weight License allows use and modification, but operating a commercial Model-as-a-Service offering requires a separate Mistral license.
  • Independent benchmark results from the earlier preview may not measure the improved October 29 weights; those results retain their original source dates.

Publisher statements, not AI IQ measurements.

About this score

How much of mistral-large-4's IQ is measured, and how the rest is estimated.

19/ 39benchmark results
6/ 7dimensions measured
1estimated
0/5Abstract
4/5Math
4/6Academic
2/6Coding
1/5Computer Use
4/6Reliability
4/6Knowledge Work
Bar height = share of benchmarks with resultsEstimatedNo score
Measured results are shown as published

Every score in the benchmark tables comes from its source unchanged. Composite IQ averages all seven dimensions equally.

Estimates fill the gaps

Abstract Reasoning has no published results yet, so its IQ is estimated from mistral-large-4's other dimensions.

A missing result is not a zero

Benchmark versions and evaluation harnesses can differ between sources, so small gaps between models may not be meaningful.

Read the scoring methodology →

Compare mistral-large-4

Side-by-side pages for the models it is most often weighed against.

mistral-large-4Comparison model
Browse all comparisons →

Frequently asked questions

Quick answers about mistral-large-4, drawn from the same data as the rest of this page.

How AI IQ scores models →Get this data from the API →
What should I know about mistral-large-4?
Mistral released improved Large 4 weights on October 29, 2026. Some independent benchmark results shown here predate that update and describe the public preview configuration.
What is the IQ of mistral-large-4?
mistral-large-4 has an estimated IQ of 118, which ranks #62 out of 186 public models on the AI IQ leaderboard. The score combines seven equally weighted reasoning dimensions using measured benchmark results and conservative estimates for missing coverage.
What is mistral-large-4 best at?
Among the dimensions with published benchmark results, mistral-large-4 ranks highest in Knowledge Work (IQ 126, #37 of 186) and Programmatic Reasoning (IQ 132, #45 of 187). Its lowest-ranked measured dimension is Computer Use (IQ 110, #97 of 186).
What is the EQ of mistral-large-4?
mistral-large-4 doesn't have an EQ score yet. EQ (emotional reasoning) is calculated from EQ-Bench 4 and AttuneBench, and neither has published a result for mistral-large-4. AI IQ doesn't estimate EQ from other benchmarks, so the score appears once one of those results is published.
Which benchmarks has mistral-large-4 been tested on?
mistral-large-4 has published results on 19 of the 39 benchmarks behind AI IQ's Composite IQ, including BullshitBench v2, AutomationBench-AA, AA Long Context Reasoning v1.1, SciCode and Terminal-Bench 4.0.
How much does mistral-large-4 cost?
mistral-large-4 costs $0.70 per million input tokens and $2.10 per million output tokens. Adjusted for how many tokens it actually consumes on real tasks, its effective cost works out to $4.69 per million input-plus-output tokens.
Is mistral-large-4 open source?
Yes — mistral-large-4 is an open-weights model: its trained weights are published, so it can be downloaded and self-hosted.
Who developed mistral-large-4?
mistral-large-4 was developed by Mistral, based in France.
When was mistral-large-4 released?
mistral-large-4 was publicly released on October 6, 2026.
How large is the context window of mistral-large-4?
mistral-large-4 supports a context window of 1,000,000 tokens — the amount of text it can consider in a single request.
How fast is mistral-large-4?
In independent measurements, mistral-large-4 generates about 116 tokens per second, with a typical response time of about 23 seconds.