Mistral: Mistral Medium 3

Mistral Medium 3, created on May 7, 2025, demonstrates a balanced and highly competitive performance profile. It exhibits competitive response times, ranking in the 49th percentile for speed across various benchmarks, indicating it performs among models with good efficiency. Its pricing is also competitive, placing in the 48th percentile, making it a cost-effective solution. Notably, the model boasts exceptional reliability, achieving the 99th percentile, signifying minimal technical failures and consistent evaluable responses. In terms of specific benchmarks, Mistral Medium 3 shows strong capabilities in Coding (87.0% accuracy, 73rd percentile) and Instruction Following (64.0% accuracy, 77th percentile). It excels in Classification tasks, achieving 99.0% accuracy in Email Classification (85th percentile). A significant strength lies in its perfect accuracy (100.0%) in both Ethics and General Knowledge benchmarks, where it is highlighted as the most accurate model at its price point and among models of comparable speed. While its Reasoning accuracy is solid at 72.0% (70th percentile), its duration for the Ethics benchmark was notably longer (19th percentile for speed). Overall, Mistral Medium 3 is a robust, reliable, and cost-efficient model with particular strengths in knowledge recall, ethical reasoning, and coding.