Mistral: Devstral Small 1.1

Mistral: Devstral Small 1.1 demonstrates a strong overall performance profile, particularly excelling in reliability and cost-efficiency. It consistently offers among the most competitive pricing, ranking in the 81st percentile across benchmarks, and exhibits exceptional reliability with a perfect 100th percentile ranking, indicating virtually no technical failures. While generally performing in the top tier for speed (69th percentile), it notably achieved a top 3 speed ranking in the Instruction Following (Baseline) benchmark. In terms of specific capabilities, Devstral Small 1.1 shows robust performance in Classification (98.0% accuracy in Email Classification) and General Knowledge (97.5% accuracy). Its core strength lies in Coding, achieving 85.0% accuracy in the Coding (Baseline) benchmark, aligning with its design for software engineering agents. A notable area for improvement is Instruction Following, where it scored 51.0% accuracy, and Reasoning, with 58.0% accuracy, placing it around the 50th percentile for these categories. Despite these areas, its high reliability and cost-effectiveness make it a compelling choice for its intended agentic coding workflows.