Qwen: Qwen3 30B A3B Thinking 2507

Qwen3-30B-A3B-Thinking-2507, a 30B parameter Mixture-of-Experts model, demonstrates strong performance in complex reasoning tasks. It consistently ranks among the fastest models and offers highly competitive pricing across all benchmarks. The model exhibits exceptional reliability with a 100% success rate, indicating robust operational stability. In terms of performance across categories, Qwen3-30B-A3B-Thinking-2507 excels in Reasoning (93rd percentile) and Ethics (100% accuracy), showcasing its advanced analytical and moral judgment capabilities. It also performs very well in Email Classification (85th percentile), Coding (78th percentile), General Knowledge (76th percentile), and Mathematics (76th percentile). A notable weakness is its Instruction Following, where it scored 0% accuracy, suggesting a significant area for improvement despite its description as having stronger instruction following. Its hallucination rate is moderate at 84% accuracy, indicating some room for improvement in acknowledging uncertainty. Overall, this model is well-suited for advanced research and agentic applications requiring structured, long-context reasoning, particularly where ethical considerations and complex problem-solving are paramount.