Qwen: Qwen3 Max

Qwen3-Max, released in September 2025, demonstrates a strong overall performance profile, particularly excelling in reliability with a perfect 100% success rate across all benchmarks, indicating exceptional stability. The model generally performs in the top tier for speed, ranking in the 63rd percentile, and offers moderate pricing, positioned at the 34th percentile. A key strength of Qwen3-Max is its remarkable accuracy in critical areas. It achieved perfect scores (100%) in Hallucinations, General Knowledge, Ethics, and Email Classification, often being the most accurate model at its price point and speed for these categories. This highlights its robust knowledge base and ability to avoid generating fabricated information. While its Mathematics, Reasoning, and Coding scores are strong (93.7%, 88.0%, and 90.6% respectively), they are not perfect, suggesting areas for potential refinement. A notable weakness appears in Instruction Following, where it scored 66.7%, placing it in the 74th percentile, indicating room for improvement in handling highly complex, multi-layered directives despite its description of following complex instructions more reliably. Its long context length of 256,000 tokens further enhances its utility for demanding applications.