STEMBench ↗2.501 / 4
Mistral Large 4 · STEMBench120 · mean progress / 4 · closed ↗
Mistral / MODEL EVIDENCE
Mistral Large 4
36 benchmarks · 40 observations · 38 sources · released 2026-10-06
Scores at a glance
36 benchmarks · large dot: Mistral Large 4; small dots: closest models where reported
ErdősBench ↗2.380 / 4
Mistral Large 4 · ErdősBench · mean progress / 4 · API ↗
GPQA Diamond ↗88.07%
Mistral Large 4 (high) · GPQA diamond · Best score (across scorers) · % ↗
Mistral Large 4 (high) · FrontierMath-Tiers-1-3-v2-Private · Best score (across scorers) · % ↗
Mistral Large 4 (high) · FrontierMath-Tier-4-v2-Private · Best score (across scorers) · % ↗
Mistral Large 4 (high) · OTIS Mock AIME 2024-2025 · Best score (across scorers) · % ↗
Mistral Large 4 (high) · SimpleQA Verified · Best score (across scorers) · % ↗
Mistral Large 4 (high) · Chess Puzzles · Best score (across scorers) · % ↗
Mistral Large 4 (high) · Mystery Game Puzzles · Best score (across scorers) · % ↗
FrontierCode 1.1 ↗39.43%
Mistral Large 4 (high) · FrontierCode · Main score · % ↗
Mistral Large 4 (high) · Furniture Assembly · Best score (across scorers) · % ↗
WeirdML (v3) ↗12.21%
Mistral Large 4 (high) · WeirdML v3 · Max performance · % ↗
Epoch Capabilities Index (ECI) ↗150.03 ECI
Mistral Large 4 · General ECI · 26 October 2026 fit ↗
Vals Finance Index ↗56.9 %
Mistral Large 4 · Vals Finance Index · Accuracy ↗
Vals Legal Index ↗65.61 %
Mistral Large 4 · Vals Legal Index · Accuracy ↗
Legal Research Bench ↗31.73 %
Mistral Large 4 · Legal Research Bench · All-Pass ↗
HLAB-Vals-1.1 ↗17.5 %
Mistral Large 4 · HLAB-Vals-1.1 · Task Pass Rate ↗
Mistral Large 4 · Harvey's Legal Agent Benchmark v1 · Accuracy ↗
Vals Finance Agent v2.1 ↗61.38 %
Mistral Large 4 · Finance Agent · Accuracy ↗
Mistral Large 4 · Excel Modeling Benchmark · Accuracy ↗
1 more reported result
56.21 % · Mistral Large 4 · Excel Modeling Benchmark v1 · Accuracy · Vals AI · Excel Modeling Benchmark v1 ↗
Tax Agent Bench ↗30.39 %
Mistral Large 4 · Tax Agent Bench · Accuracy ↗
Vals Finance Agent v2 ↗54.68 %
Mistral Large 4 · Finance Agent v2 · Accuracy ↗
MedCode ↗40.69 %
Mistral Large 4 · MedCode · Accuracy ↗
MedScribe ↗80.43 %
Mistral Large 4 · MedScribe · Accuracy ↗
SAFE-Teen ↗33%
Mistral Large 4 · Urgent safety · share of applicable ratings marked fail ↗
2 more reported results
53% · Mistral Large 4 · Health advice · share of applicable ratings marked fail · Vals AI · SAFE-Teen ↗
19% · Mistral Large 4 · Relationship boundaries · share of applicable ratings marked fail · Vals AI · SAFE-Teen ↗
ProofBench ↗10 %
Mistral Large 4 · ProofBench · Accuracy ↗
BioMysteryBench ↗67.04 %
Mistral Large 4 · BioMysteryBench · Accuracy ↗
MysteryMechanism ↗12.16 %
Mistral Large 4 · MysteryMechanism · Accuracy ↗
Vibe Code Bench ↗14.26 %
Mistral Large 4 · Vibe Code Bench · Vibe Code Bench 1–100 ↗
1 more reported result
78.4 % · Mistral Large 4 (OpenHands) · Vibe Code Bench · original task set · Vals AI · Vibe Code Bench ↗
Terminal-Bench 4.0 ↗27.78 %
Mistral Large 4 (Mini-SWE-agent) · Terminal-Bench 4.0 · Score ↗
IOI v1 & IOI v2 ↗45.28 %
Mistral Large 4 · IOI · Accuracy ↗
SWE-bench Multimodal ↗33.13 %
Mistral Large 4 · SWE-bench Multimodal · Accuracy ↗
Code Migration v1.1 ↗30.56 %
Mistral Large 4 · Code Migration v1 · Accuracy ↗
DistressBench ↗77.02%
mistral-large-4 · DistressBench · Score ↗
Fortress ↗23.07%
mistral-large-4 · Scale Labs · Risk elicitation score ↗
ALE-Bench ↗1213.15 Performance points
mistral-large-4 · Average contest performance ↗
Closest models by score
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
Claude Sonnet 5 →Anthropic
18 of 30 shared benchmark comparisons
DeepSeek V4.1 Flash →DeepSeek
25 of 30 shared benchmark comparisons
DeepSeek V4 Pro 0813 →DeepSeek
22 of 30 shared benchmark comparisons
How similarity is measured
Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.
Sources
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
- FrontierCode Leaderboard ↗
- Ale Bench — methodology and published results ↗
- Chess Puzzles ↗
- FrontierMath Tier 4 (v2) ↗
- FrontierMath Tiers 1-3 (v2) ↗
- Furniture Assembly Benchmark ↗
- GPQA Diamond ↗
- Mystery Game Puzzles ↗
- OTIS Mock AIME 2024-2025 ↗
- SimpleQA Verified ↗
- AI benchmarks & capabilities ↗
- WeirdML ↗
- DistressBench — evaluation methodology and results ↗
- Scale Labs — evaluation methodology and results ↗
- ErdősBench: research mathematics, scoped progress and proof audits ↗
- STEMBench — Six-Domain Model Evaluation | Ulam ↗
- BioMysteryBench — evaluation methodology and results ↗
- Code Migration v1 — evaluation methodology and results ↗
- Excel Modeling Benchmark v1 — evaluation methodology and results ↗
- Excel Modeling Benchmark — evaluation methodology and results ↗
- Finance Agent v2 — evaluation methodology and results ↗
- Finance Agent — evaluation methodology and results ↗
- Vals Finance Index — evaluation methodology and results ↗
- HLAB-Vals-1.1 — evaluation methodology and results ↗
- Harvey's Legal Agent Benchmark v1 — evaluation methodology and results ↗
- IOI — evaluation methodology and results ↗
- Vals Legal Index — evaluation methodology and results ↗
- Legal Research Bench — evaluation methodology and results ↗
- MedCode — evaluation methodology and results ↗
- MedScribe — evaluation methodology and results ↗
- SAFE-Teen: Teen Mental Health & Safety | Vals AI A conversation branches after the chatbot turns down a request — evaluation methodology and results ↗
- MysteryMechanism — evaluation methodology and results ↗
- ProofBench — evaluation methodology and results ↗
- SWE-bench Multimodal — evaluation methodology and results ↗
- Tax Agent Bench — evaluation methodology and results ↗
- Terminal-Bench 4.0 — evaluation methodology and results ↗
- Vibe Code Bench 1-100 — evaluation methodology and results ↗
- Vibe Code Bench — evaluation methodology and results ↗
Names used in source tables
Mistral Large 4 · Mistral Large 4 (Mini-SWE-agent) · Mistral Large 4 (OpenHands) · Mistral Large 4 (high) · mistral-large-4