Mistral Medium 3 delivers frontier performance while being an order of magnitude less expensive. For instance, the model performs at or above 90% of Claude Sonnet 3.7 on benchmarks across the board at a significantly lower cost. On performance, Mistral Medium 3 also surpasses leading open models such as Llama 4 Maverick and enterprise models such as Cohere Command A. On pricing, the model beats cost leaders such as DeepSeek v3, both in API and self-deployed systems.
Added Sep 25, 2025
Context Window
131.1K
Max Output
32.8K
Input Price (Auto)
$0.40/1M
Output Price (Auto)
$2.00/1M
Performance metrics and benchmarks
Sourced from Artificial Analysis.
Intelligence Index
18.8
Auto routing is available for this model. Explicit provider selection is not available.
Loading provider options…
Coding Index
13.6
Agentic Index
13.7
GPQA Diamond
Graduate-level scientific reasoning
57.8%
Better than 44% of models compared
HLE
Humanity's Last Exam
4.3%
Better than 23% of models compared
IFBench
Instruction-following benchmark
39.3%
Better than 44% of models compared
T²-Bench Telecom
Conversational AI agents in dual-control scenarios
24.3%
Better than 32% of models compared
AA-LCR
Long context reasoning evaluation
28.0%
Better than 47% of models compared
GDPval-AA
Economically valuable tasks
4.4%
Better than 47% of models compared
CritPt
Research-level physics reasoning
0.0%
Better than 36% of models compared
SciCode
Python programming for scientific computing
33.1%
Better than 59% of models compared
Terminal-Bench Hard
Agentic coding and terminal use
3.8%
Better than 28% of models compared
AIME 2025
American Invitational Mathematics Examination 2025
30.3%
Better than 32% of models compared
AIME
American Invitational Mathematics Examination
44.0%
Better than 66% of models compared
MMLU-Pro
Professional and academic subject knowledge
76.0%
Better than 52% of models compared
AA-Omniscience Accuracy
Proportion of correctly answered questions
18.3%
Better than 53% of models compared
Last updated May 15, 2026
Artificial AnalysisLiveCodeBench
Contamination-free coding benchmark
40.0%
Better than 47% of models compared
Math-500
Diverse mathematical problem solving benchmark
90.7%
Better than 65% of models compared
AA-Omniscience Hallucination Rate
Rate of incorrect answers among non-correct responses
60.9%
Better than 90% of models compared
