阿里(09988)通义千问推出新模型Qwen3-30B-A3B-Thinking-2507

智通财经
Jul 31, 2025

智通财经APP获悉,7月31日,阿里(09988)通义千问向大家介绍新模型Qwen3-30B-A3B-Thinking-2507,一个更睿智、更敏捷、更全能的新推理模型。目前,Qwen3-30B-A3B-Thinking-2507已在魔搭社区、HuggingFace上开源。

相较于4月29日开源的Qwen3-30-A3B模型,新模型在推理能力、通用能力及上下文长度上有了显著提升:

数学超强,代码贼溜:新模型在聚焦数学能力的AIME25评测中斩获85.0的高分,在代码能力测试LiveCodeBench v6中得分66.0,两项核心推理能力均超越Gemini2.5-Flash(thinking)、Qwen3-235B-A22B(thinking);新模型的知识水平(GPQA、MMLU-Pro)也较上一版本进步显著。

善解人意,全面升级:在写作(WritingBench)、Agent能力(BFCL-v3)、多轮对话和多语言指令遵循(MultiIF)等通用能力评测中,Qwen3-30B-A3B-Thinking-2507均超越Gemini2.5-Flash(thinking)、Qwen3-235B-A22B(thinking)。

更长的上下文理解:原生支持256K tokens,可扩展至1M tokens。

此外,新模型的思考长度(thinking length)也增加了,推荐在高度复杂的推理任务中设置更长的思考预算,以充分发挥它的潜力。

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10