Mistral: Ministral 3B

Mistral: Ministral 3B, created on October 16, 2024, by mistralai, is a 3B parameter model optimized for on-device and edge computing, supporting an impressive 128k context length. This model consistently ranks among the fastest, achieving the 89th percentile across six benchmarks, and offers highly competitive pricing, placing in the 95th percentile. Its reliability is exceptional, demonstrating minimal technical failures and ranking in the 100th percentile. While excelling in speed, cost, and reliability, Ministral 3B exhibits a mixed performance across specific benchmarks. It shows strong capabilities in General Knowledge (90.5% accuracy, 35th percentile), notably being the most accurate model at its price point, and Ethics (96.0% accuracy, 32nd percentile). However, its performance in Reasoning (14.0% accuracy, 10th percentile) and Instruction Following (31.0% accuracy, 31st percentile) is a notable weakness. Coding (75.0% accuracy, 41st percentile) and Email Classification (92.0% accuracy, 24th percentile) show moderate accuracy but are not standout strengths. Ministral 3B's primary strengths lie in its cost-effectiveness, rapid inference, and robust reliability, making it ideal for resource-constrained environments and agentic workflows where these factors are paramount, despite some limitations in complex reasoning tasks.