Artificial Analysis benchmarks Mistral Large 4 open weights, scores 26 on Intelligence Index
Article timeline
The articles in this story. Times are ET.
Artificial AnalysisOfficialArtificial Analysis links full results for Mistral Large 4AIArtificial Analysis links to its full benchmark results for Mistral Large 4 on its model page. The post itself gives no scores, prices, or other figures.
- Artificial AnalysisOfficialArtificial Analysis publishes Intelligence Index results for Mistral Large 4
AIArtificial Analysis has published full Intelligence Index results for Mistral Large 4 at high reasoning effort. The post shares the complete index results but gives no scores or other figures in the text provided.
- Artificial AnalysisOfficialMistral Large 4 ranks second among open-weights models on coding agent index
AIRunning in Mistral Vibe, Mistral Large 4 (high) scores 48 on the Artificial Analysis Coding Agent Index, placing second among current open-weights models. It uses about 8.5M total tokens per task, roughly half the 15.9M and 16.7M used by GLM-5.3 (max) in Opencode and DeepSeek V4.1 Flash 1017 (max) in DeepSeek Harness.
- Artificial AnalysisOfficialMistral Large 4 (high) uses fewer output tokens than higher-scoring open models
AIArtificial Analysis says Mistral Large 4 (high) uses 82K output tokens per Intelligence Index task, sitting on the open weights Pareto frontier. Open weights models that score higher use 113K or more output tokens per task. Compared to the Preview, Mistral Large 4 (high) uses 39% fewer output tokens per Intelligence Index task.
- Artificial AnalysisOfficialPickMistral Large 4 open weights scores 26 on Artificial Analysis Intelligence Index
AIArtificial Analysis says Mistral has released Mistral Large 4 open weights, scoring 26 on its Intelligence Index, narrowly the top non-Chinese open-weights model. The model has 1.05T total and 52B active parameters, uses 39% fewer output tokens than its Preview, and costs $0.70/$2.10 per 1M input/output tokens.
- Artificial AnalysisOfficialMistral Large 4 ranks third among open weights models on Cyber Index
AIMistral Large 4 (high) scores 51 on the Artificial Analysis Cyber Index, third among current open weights models behind DeepSeek V4.1 Flash 1017 (max, 57) and MiMo-V2.6-Pro (56). It matches or beats both on CyberGym-E2E-AA at 79%, against 76% and 79%, but trails them on DeepsecBench-AA (15% vs. 25% and 26%) and CWE-Bench-AA (58% vs. 69% and 63%).
