Alibaba Announces Qwen3.8: A 2.4T Parameter Open-Weight Powerhouse

Alibaba has announced Qwen3.8, a 2.4T parameter model that will soon be released as open-weight. It is currently ranked as a top-tier model, second only to Fable 5 in performance. Users can immediately test the preview version through Alibaba's Token Plan and Qoder platforms.

Key Points

• Alibaba is launching Qwen3.8, a 2.4 trillion parameter model that will eventually be open-weight.
• The model is claimed to be one of the most powerful available, ranking second only to Fable 5.
• A preview version, Qwen3.8-Max-Preview, is immediately available for testing on Alibaba's platforms.
• The community is showing high engagement, with specific requests for smaller versions to facilitate local deployment.

Sentiment

The overall sentiment is cautiously positive toward the technical and competitive significance of Qwen3.8, with clear enthusiasm for open weights and skepticism toward closed Western labs. At the same time, the thread is highly mixed because many commenters attach concerns about censorship, benchmark reliability, national strategy, and the practical usability of very large models. HN does not simply endorse the announcement, but it largely treats it as important and credible enough to intensify the open-versus-closed model debate.

In Agreement

• Several commenters see Qwen3.8 as evidence that open-weight models are rapidly closing the gap with closed frontier systems and forcing Western labs to justify their closed releases.
• Users familiar with prior Qwen releases argue that Alibaba's models are already strong in practice, especially when judged by API quality, price, and availability rather than brand reputation.
• Some commenters welcome Chinese labs' willingness to publish weights, saying it benefits developers, researchers, and hobbyists who want more control than subscription-only model access provides.
• A number of participants frame the release as part of a broader commoditization of frontier AI, where model capability becomes less defensible as more labs can train competitive systems.
• Technical commenters note that even extremely large open-weight models can be useful locally in constrained or specialized workflows, especially if only a subset of parameters is active or if slow inference is acceptable for high-value tasks.

Opposed

• Skeptics question whether the model's claimed benchmark position reflects durable real-world quality, noting that public benchmarks can be gamed or overfit and that preview access is not the same as a full release.
• Some commenters doubt whether Chinese open-weight models are always trained independently, raising concerns about distillation from closed Western systems or other opaque training practices.
• A recurring concern is censorship and political alignment: several participants argue that Chinese models may be constrained on sensitive topics or shaped by state interests in ways that matter for trust.
• Others push back on the practical value of a model at this scale, arguing that open weights are less useful if ordinary developers cannot run them at reasonable speed or cost.
• A substantial side of the thread argues that geopolitical context cannot be separated from Chinese AI companies because of perceived ties between large firms and the Chinese state.