Frontier BenchAI model benchmarks · Jul 2026
ModelsCompareLabsMethodology
All models
MA

Mistral Small 4

Mistral AI

119B dense. Runs on a single 8×H100 node — easy self-host generalist.

Open-weightApache 2.0128K contextReleased Apr 2026textcode

API price · per 1M tokens

$0.1/$0.3

input / output

≈ $0.15/M blended

Benchmarks

Bar shows this model's score; tick shows the best score across all tracked models for context.

SWE-bench VerifiedReal GitHub issues, human-verified patches. The standard coding-agent benchmark.
58.0/ best 90.1
GPQA DiamondGoogle-proof graduate-level science Q&A. General reasoning proxy.
70.0/ best 94.4
MMLU-ProMulti-task language understanding across 120+ academic subjects.
72.0/ best 93.0
HumanEval+Functional code generation from docstrings.
85.0/ best 96.0
MATH-500Competition-style math problems.
80.0/ best 97.8

Other Mistral AI models

MA

Mistral Large 3

Open-weight

≈$0.60/M

Built by GLM-5.2 · Scores are directional, drawn from public mid-2026 benchmarks.

SWE-bench, GPQA, MMLU, AIME, HLE etc. are trademarks of their respective owners.