Alibaba Previews Qwen3.8 Model with 2.4 Trillion Parameters, Weights to Be Released
Alibaba's Qwen team has announced a preview of its upcoming Qwen3.8 large language model, boasting an impressive 2.4 trillion parameters. The company stated that the model's weights are slated for release in the near future. A preliminary version, identified as Qwen3.8-Max-Preview, is currently accessible to users through Alibaba's Token Plan, Qoder, and QoderWork platforms. However, Alibaba has yet to disclose specific details regarding the model's internal architecture, the datasets used for its training, its performance metrics based on benchmark tests, or a definitive timeline for its full release.
The unveiling of Qwen3.8 with 2.4 trillion parameters signifies a significant escalation in the scale of foundational models, potentially impacting the competitive landscape of AI development. While the release of weights could foster broader research and application, the lack of detailed architectural and training data information presents challenges for independent verification and understanding of potential biases or limitations. The strategic decision to preview and gradually release access suggests a phased market entry, allowing Alibaba to gauge adoption and refine the model based on early user feedback. This approach may also be influenced by the evolving regulatory environment surrounding AI, prompting a cautious rollout of highly capable systems.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.