Flash News: Qwen 3.8 Is Going Open-Weight

Qwen 3.8 Is Going Open-Weight. If You're Still Calling Kimi K3 a Fluke, You May Want to Reconsider.

Just days after I argued that Kimi K3 was more than just another model release, we now have another strong signal pointing in the same direction: Qwen 3.8 is reportedly arriving as an open-weight model.

So far, Alibaba has only announced the flagship 2.7T-parameter model. Yet the community reaction has been overwhelmingly positive. Why?

Much of that excitement comes from the expectation that Alibaba won't stop at the flagship. Many in the community expect a smaller model (in the 30B parameter range), following the successful Qwen 3.6 lineup. While nothing has been confirmed, the possibility alone has sparked a wave of discussion because it could take model running on consumer hardware to another level.

More importantly, this isn't just about Qwen. It reinforces the broader trend I argued after Kimi K3: open-weight frontier models are no longer isolated surprises. They are arriving with increasing frequency, and the performance gap with the best closed models is narrowing much faster than conventional wisdom predicted.

If that trend continues, the competitive advantage of keeping frontier models closed becomes harder to justify. Labs such as Anthropic may increasingly need to compete on product, ecosystem, and services instead of hanging on just model quality.

➡️ Read the full analysis: Why Kimi K3 Is More Than Another Model Release

FAQ

Can I run Kimi K3, GLM, or the Qwen 3.8 flagship on my own computer?

No. Models such as Kimi K3, GLM, and the announced Qwen 3.8 2.7T are frontier-scale models that require multiple high-memory GPUs or server-class AI accelerators to run efficiently. They are well beyond the capabilities of a typical desktop PC.

That said, model families rarely consist of a single model. Previous releases have included much smaller variants that preserve many of the flagship's capabilities while being practical on consumer hardware.

Has Alibaba confirmed a smaller Qwen 3.8 model?

No. At the time of writing, Alibaba has only announced the 2.7T-parameter model. The speculation about a smaller variant comes from the community, which expects Alibaba to follow the same strategy used with previous Qwen releases. Any details about a 30B-class model remain speculation.

Does this mean open models have already surpassed closed models?

The very best closed models still lead on many benchmarks and capabilities. However, recent releases such as Kimi K3, GLM, and now Qwen 3.8 suggest that the gap is shrinking much faster than many expected. Instead of a two-year lag between open and closed models, we're increasingly seeing open-weight models approach frontier performance within months.