← Founder Notes
Archive

The 2.4-trillion model just went open-weight. alibaba released qwen3.8-max in october under its own…

Yethikrishna ROriginal on Threads

the 2.4-trillion model just went open-weight. alibaba released qwen3.8-max in october under its own license, the biggest open-weight model published to date.

most teams can now download a frontier they can't run.

Context

Verified dates: Alibaba announced Qwen3.8-Max on Aug 2 to 3, 2026 (Qwen blog; Alibaba Group), with 2.4 trillion parameters and a context window of up to 1 million tokens, and said it was the first Max-class Qwen to get open weights. The weights landed on Hugging Face on Aug 12, 2026 as Qwen3.8-2.4T-A95B (explainx; ZAVINO), with 95 billion active parameters.

Backgrind (Aug 23) says the download is 4.89 TB in 224 files with a license catch, and explainx describes a new revenue-share license and a text-only checkpoint.

How it compares

The post says Alibaba released the model in October. The announcement was Aug 2 to 3 and the open weights were published Aug 12, so October is not the release date; an Oct 5 article (AI Intel Report) is later coverage. The 2.4 trillion parameter size matches. 'The biggest open-weight model published to date' was not seen in the sources read, so unsupported here, not refuted. Backgrind compares the download with Kimi K3 in a way that suggests Kimi K3 is larger by parameter count. 'Most teams can download a frontier they cannot run' is the author's opinion; ZAVINO makes the same point about hardware.

Related work

Watch next

  • Read the license text on Hugging Face and the Kimi K3 parameter count to settle 'biggest'.

Sources

  1. Qwen blog: Qwen3.8-Max, a new bar for coding and coworkqwen.ai
  2. explainx: Qwen3.8-Max open weights are liveexplainx.ai
  3. Backgrind: Qwen3.8-Max open weights, 4.89 TB, 224 files, one license catchbackgrind.com

Provenance

The note above is reproduced unedited from the original post, first published on Threads on 11 October 2026 at 16:54 IST. Sources are the papers and datasets the note draws on.

View the original post
Embed this note
<iframe src="https://founder.myndlabs.tech/notes/embed/the-2-4-trillion-model-just-went-open-DeWi0PGDFFh" width="480" height="420" style="border:0;max-width:100%" loading="lazy" title="The 2.4-trillion model just went open-weight. alibaba released qwen3.8-max in october under its own…"></iframe>

More notes