Qwen 3.6

weights

open(Apache-2.0 for open variants e.g. 27B/35B-A3B)

data

closed

code

partial(inference/usage code on GitHub

license

Apache-2.0(most restrictive distributed SKU

Apache-2.0 across the distributed base variants, with a closed corpus and only inference code published, so 3. The exclusion that carries the score: Qwen3.7-Max is proprietary and served only through Alibaba Model Studio, so it is never distributed and does not cap the family. Counting it would publish Qwen as closed, which would say nothing useful about the weights anyone can actually download - an API-only tier is a different product surface. Qwen 2.5 scored 2, because its distributed 3B and 72B checkpoints carried the Qwen License Agreement; from Qwen3 onward every distributed variant is Apache-2.0.

• https://huggingface.co/api/models/Qwen/Qwen3.6-27B recorded 2026-08-13

  card license `apache-2.0`, `gated: false`, `private: false`, 6,830,615 downloads and 2,252 likes; the card front matter declares no `datasets`, so no corpus is named there.

• https://api.github.com/repos/QwenLM/Qwen3.6 recorded 2026-08-13

  `private: false`, `archived: false`, license `Apache-2.0`, description "Qwen3.6 is the large language model series developed by Qwen team, Alibaba Group". Usage and deployment code, no pretraining pipeline, which is what holds `code` at partial.

• https://github.com/QwenLM/Qwen3.6 recorded 2026-08-15

  Repository page for QwenLM/Qwen3.6: 'Apache-2.0 license' in the sidebar, 3.8k stars, public and unarchived. Establishes the Apache-2.0 tier across the released base variants and that the published code is inference-side.

• https://www.marktechpost.com/2026/04/22/alibaba-qwen-team-releases-qwen3-6-27b-a-dense-open-weight-model-outperforming-397b-moe-on-agentic-coding-benchmarks/ recorded 2026-08-15

  Trade coverage of the Qwen3.6-27B open-weight release and its 262K context. Corroborates the open-weights reading; establishes no dimension on its own.

• https://llm-stats.com/models/qwen3.6-plus recorded 2026-08-15

  NOT READABLE: the page returns HTTP 200 and serves a 'Quick verification - Confirm you're human to keep going' interstitial rather than the model page, so it shows nothing and establishes nothing. Retained because the record cites it; a digest of the interstitial is recorded rather than a claim about the model.

• https://aimlapi.com/blog/qwen-3-6-series-alibabas-open-source-llm-revolution-in-2026 recorded 2026-08-15

  Vendor-blog overview of the Qwen3.6 series describing it as open-weight with a 128K context. Third-party commentary; establishes no dimension on its own.

14,030,698 downloads in the trailing 30 days across the six declared artifacts (Qwen/Qwen2.5-72B 28,852; Qwen/Qwen2.5-7B 716,344; Qwen/Qwen3.6-27B 6,830,615; Qwen/Qwen3.6-35B-A3B 5,609,980; Qwen/Qwen3-235B-A22B 529,067; Qwen/Qwen3.5-397B-A17B 315,840), which bands at level 5 (>10M) on the model adoption scale. The two Qwen3.6 checkpoints supply 89% of that total, so the family's reach is current rather than inherited from the 2.5 generation.

• https://huggingface.co/api/models/Qwen/Qwen2.5-72B recorded 2026-08-13

  28,852 downloads in the trailing 30 days for Qwen/Qwen2.5-72B

• https://huggingface.co/api/models/Qwen/Qwen2.5-7B recorded 2026-08-13

  716,344 downloads in the trailing 30 days for Qwen/Qwen2.5-7B

• https://huggingface.co/api/models/Qwen/Qwen3.6-27B recorded 2026-08-13

  6,830,615 downloads in the trailing 30 days for Qwen/Qwen3.6-27B

• https://huggingface.co/api/models/Qwen/Qwen3.6-35B-A3B recorded 2026-08-13

  5,609,980 downloads in the trailing 30 days for Qwen/Qwen3.6-35B-A3B

• https://huggingface.co/api/models/Qwen/Qwen3-235B-A22B recorded 2026-08-13

  529,067 downloads in the trailing 30 days for Qwen/Qwen3-235B-A22B

• https://huggingface.co/api/models/Qwen/Qwen3.5-397B-A17B recorded 2026-08-13

  315,840 downloads in the trailing 30 days for Qwen/Qwen3.5-397B-A17B

Dense Qwen3.6-27B: SWE-bench Verified 77.2, Terminal-Bench 2.0 59.3 (reported matching Claude 4.5 Opus class), outperforming a 397B MoE on agentic coding. Flagship Qwen3.6-Plus reported matching Claude Opus 4.5 on SWE-bench/Terminal-Bench. Frontier-tier open family.

• https://www.marktechpost.com/2026/04/22/alibaba-qwen-team-releases-qwen3-6-27b-a-dense-open-weight-model-outperforming-397b-moe-on-agentic-coding-benchmarks/ recorded 2026-08-13

  Qwen3.6-27B SWE-bench Verified 77.2 and Terminal-Bench 2.0 59.3, outperforming a 397B MoE on agentic coding

• https://www.caixinglobal.com/2026-04-02/alibaba-releases-qwen-36-plus-ai-model-with-enhanced-coding-capabilities-102430395.html recorded 2026-08-13

  "Alibaba Group Holding Ltd. released Qwen 3.6-Plus, a multimodal LLM excelling in agentic coding ... rivals Claude Opus 4.5 on SWE-bench"