Qwen: Qwen3 VL 30B A3B Thinking

Qwen3 VL 30B A3B Thinking demonstrates moderate speed performance, ranking in the 20th percentile, and offers moderate pricing, placing it in the 39th percentile. A standout feature is its exceptional reliability, achieving a 99% success rate across benchmarks, indicating minimal technical failures. In terms of performance across categories, the model excels in Ethics, achieving perfect accuracy and being noted as the most accurate at its price point and speed. It also shows strong capabilities in Email Classification (99.0% accuracy) and Coding (90.7% accuracy), performing well above average. Its "Thinking" variant's focus on reasoning is reflected in a respectable 67.3% accuracy in the Reasoning benchmark and 91.9% in Mathematics. However, the model exhibits notable weaknesses in Instruction Following, with a low 25.3% accuracy, and struggles with Hallucinations, achieving only 84.0% accuracy, suggesting it sometimes fails to appropriately acknowledge uncertainty. General Knowledge is solid at 97.0%. Overall, Qwen3 VL 30B A3B Thinking is a reliable multimodal model with strong ethical understanding and coding abilities, though it could improve in instruction adherence and hallucination mitigation.