Qwen: Qwen3 235B A22B

Qwen3-235B-A22B, a 235B parameter MoE model, exhibits exceptional reliability with a 100% success rate across all benchmarks, consistently providing usable responses. However, its speed performance is a notable area for improvement, tending to have longer response times, ranking in the 10th percentile. Pricing is moderate, falling into the 38th percentile. The model demonstrates significant strengths in several key areas. It achieved perfect accuracy in both General Knowledge and Email Classification, with the latter also being the most accurate and fastest among models at its price point. Its Coding capabilities are outstanding, scoring 97.0% accuracy (97th percentile), and it shows strong Reasoning abilities at 98.0% accuracy (91st percentile). Ethics performance is also solid at 99.0%. The primary weakness lies in Instruction Following, where it achieved only 40.4% accuracy (32nd percentile), indicating challenges with complex, multi-layered instructions. Despite its "thinking" mode for complex tasks, the model's overall speed remains a bottleneck.