Qwen3-2507Open Weight

Released

Qwen3-2507 is an AI model released by Qwen on Monday, Jul 21 2025, 83 days after Qwen3. It is an open-weight model — the trained weights are available to download and run. It was released in 4B and 30B and 235B parameter sizesThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality., with a 256k token context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider.. Benchmark results (shown below) cover LiveCodeBench, IFEval, Humanity's Last Exam, AIME 2025, GPQA Diamond, and MMLU-Pro.

Benchmarks

Coding

LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better.
74.1%
#30 of 86

Agentic & tool use

IFEvalAgentic tasks — following verifiable instructions in generated responses
87.8%
#7 of 24

Reasoning & science

Humanity's Last ExamMultidisciplinary reasoning — accuracy on the 2,158 text-only questions without tools
18.2%text only, no tools
#16 of 27
AIME 2025Reasoning — single-answer accuracy on the 2025 invitational mathematics exam without tools
92.3%
#13 of 47
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.
81.1%
#65 of 145

Knowledge work

MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark
84.4%
#16 of 58

Compare Qwen3-2507 with

Qwen3-2507

About

Qwen3-2507 began on July 21, 2025 with a post-trained 235B Instruct model. Its companion 235B Thinking checkpoint followed on July 25, and smaller 30B and 4B sizes joined the same refresh by August 6.

This release separated direct responses from extended reasoning, which the April Qwen3 models had supported in the same checkpoint. The 235B Thinking model scored 92.3 on AIME 2025 and 74.1 on LiveCodeBench v6 in Qwen's model card. Qwen later documented a context extension to roughly one million tokens for the larger 2507 checkpoints.

Frequently asked questions

Qwen3-2507 was released by Qwen on Monday, Jul 21 2025.

All Qwen releases

42 tracked