AI BriefingQwen to Launch 27B Model Soon
·2026.08.05 20:51
Key point
The Qwen development team teased the launch of a 27B model and key specifications for upcoming models.
Details
The Qwen development team revealed plans for upcoming models and product updates during a Twitter/X AMA.
- They announced that a 27B model will be released soon, aiming for a new level of performance rather than a simple retraining of previous models.
- Qwen 3.8 was introduced with a total of 2.4T parameters and 95B active parameters.
- They explained that the capability to understand videos longer than 100 hours is closer to a hierarchical video memory system than to traditional agent swarms. It stores scenes, entities, events, and temporal relationships from videos as structured text graphs for use in retrieval and reasoning.
- Regarding code development tools, they teased further updates to Qoder and QwenWork.
- They did not disclose specific figures for pre-training data composition or reinforcement learning scale, stating only that large-scale computing was used.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Apple Releases LensVLM-9B for Processing Compressed Text Images