Benchmark

Mistral — reported benchmarks

Mistral AI · source page ↗ · last checked Oct 8, 2026, 6:03 AM

Reported benchmarks · Mistral Large 4

captured Oct 8, 2026, 6:03 AM
BenchmarkScore
DeepSWE v1.161.7%
SWE-Atlas-QnA59.4%
Terminal-Bench 428.3%
Coding Agent Index49.8%
AutomationBench59.9% · 657 business workflows across Gmail, Google Sheets, Slack, Salesforce
Cybench93% · 40 security challenges
CyberGym-E2E-AA82% · Reproduce and patch vulnerability test
Lakera B3 Attack Resistance93.3%
AA-Briefcase1393 Elo
Dense20042% · Visual grounding benchmark

Vendor-reported via automated web search — not independently verified. See the cited matrix on /models.

In the news · Mistral AI

Importance-filtered press coverage (Google News) mentioning Mistral AI. Headlines link to the original; verify before acting.

Change history