Chinese LLMs See Intensive Updates
智东西 · wechat · 2026-07-19
This article reviews the recent wave of intensive updates from Chinese LLMs. First, Alibaba's Qwen officially announced that Qwen3.8 is coming soon and will be open-sourced with a parameter scale of 2.4T. A preview version of Qwen3.8-Max is already live on Alibaba TokenPlan, Qoder, and QoderWork, accompanied by limited-time offers.
The article also mentions that Kimi-K3 has been open-sourced. With 2.8T parameters, it has won praise from overseas developers for its long-range task handling, frontend rendering, and AgentSwarm capabilities. Additionally, the official version of DeepSeek-V4 is reportedly in grey-box testing among some users, and MiniMax M3 Pro is said to be launching soon with plans to go open source. Overall, it captures the current "blowout" phase of domestic Chinese models.
Related event: Alibaba Announces 2.4T Open-Weight Model Qwen3.8(22 posts)→
More from Models
- Cohere's Invent-A-Dataset paper: zero-seed synthetic data beats frontier APIs on quality and diversity — idanbeck · 2026-10-27
- ARC Prize redesigns leaderboard: highest reasoning effort by default, CSV export — mhmazur · 2026-10-27
- Cohere: open-weight Qwen3.8-Flash-Next tried uploading probe file to Beijing cloud bucket in sandbox — wstone_bd · 2026-10-27
- Grok Bot auto-connects to X Money accounts without opt-in, user pulls funds — noahsolomon · 2026-10-27
- Model refuses wrong psychiatric advice in deliberately misleading test — FellMentKE · 2026-10-27
- Clinical System One v0.2: top clinical AI drops from 97% to 65% on harder cases — MaziyarPanahi · 2026-10-27