Acropolis

Mistral Large 4 OSS

MistralGrade B. Evidence grade B of A to D: good.Open weightsReleased 29 Oct 2026Consistent across benchmarks
Compare with peers

Domain profile

Median within-board percentile by capability domain across boards with sufficient peer overlap.

  • Agentic (9 boards): 43 (scale 0 to 100)Agentic (9 boards)43
  • Coding (4 boards): 43 (scale 0 to 100)Coding (4 boards)43
  • Finance (4 boards): 41 (scale 0 to 100)Finance (4 boards)41
  • Legal (3 boards): 43 (scale 0 to 100)Legal (3 boards)43
  • Science (1 boards): 79 (scale 0 to 100)Science (1 boards)79

Benchmark appearances

Best verified run per board, with confidence intervals and Pareto frontier status.

BenchmarkRankScoreCostStatusObserved
Legal Agent Benchmark (HLAB-Vals)

Vals AI benchmarks · high

#10 / 38
15.8%[9.7% to 22%]$0.73Standard15 Nov 2026
Code Migration

Vals AI benchmarks · high

#20 / 38
51.3%[42.6% to 60.1%]$11.77Standard15 Nov 2026
Finance Agent

Vals AI benchmarks · high

#20 / 38
61.7%[57.7% to 65.7%]$0.35Standard15 Nov 2026
Vals Index

Vals AI benchmarks · high

#22 / 38
56.5%[54.1% to 59%]$3.72Standard15 Nov 2026
Vibe Code Bench

Vals AI benchmarks · high · OpenHands

#22 / 38
81.6%[75.1% to 88%]$11.72Standard15 Nov 2026
Tax Agent

Vals AI benchmarks · high

#24 / 38
30.7%[24.2% to 37.1%]$0.50Standard15 Nov 2026
Terminal-Bench 4

Vals AI benchmarks · high

#24 / 38
26.3%[16.2% to 36.3%]$2.73Standard15 Nov 2026
EMB

Vals AI benchmarks · high

#26 / 38
68.7%[63.9% to 73.4%]$0.83Standard15 Nov 2026
Legal Research

Vals AI benchmarks · high

#30 / 38
35.6%[29.1% to 42.1%]$1.09Standard15 Nov 2026
GPQA diamond

Epoch AI Benchmarking Hub · high

#54 / 338
90.2%[86.6% to 93.7%]UnknownStandard2 Nov 2026

Closest head-to-head rivals

Direct pairwise win rates for Mistral Large 4 OSS across shared benchmark boards.