Qwen3-30B-A3B-Thinking-2507
-
Qwen3 New Inference Models Released: Significantly Enhanced Reasoning, Math, and Code Capabilities
Alibaba’s Tongyi Qianwen introduced Qwen3-30B-A3B-Thinking-2507, a revamped reasoning model with significant upgrades in math and coding. It outperforms competitors like Gemini 2.5-Flash on benchmarks like AIME25 and LiveCodeBench v6. The model also shows improved performance in knowledge assessments, writing, agent functionality, and multi-lingual instruction understanding. Its expanded “thinking length” allows for more complex problem-solving. The model and Qwen Chat platform are available on open-source platforms, providing developers with a robust AI tool.
