Symbolic imageAlibaba releases largest Qwen model with open weights
Alibaba unveiled Qwen3.8-Max, its most capable AI model to date, on Monday and made its weights downloadable. A research firm put DeepSeek's new V4-Flash at more than 100 times cheaper than Anthropic's Claude Fable 5.
What happened
- Qwen3.8-Max has 2.4 trillion parameters, behind Moonshot AI's Kimi K3, launched last month with 2.8 trillion.
- Alibaba shares rose 7% in Hong Kong trade after the launch on Monday, 3 August.
- On Arena.AI it became the highest-ranking Chinese text model, behind Anthropic's Claude Fable 5 and three Opus variants.
- On Arena.AI's leaderboard for visual material, Qwen3.8-Max ranked second globally, behind a Claude Fable 5 variant.
- Qwen3.8-Max and Kimi K3 process text, images and video, up to 1 million tokens at once.
The view from outside
Daily Sabah, carrying the Reuters wire from outside the US-China rivalry, foregrounds Omdia analyst Lian Jye Su's argument that most business workflows do not need the industry's best model, only one that is affordable, transparent and accessible.
What could happen next
- Open whether downloadable weights convert into sustained developer adoption abroad or remain leaderboard placements.
- Global Times reports MiniMax will also open-source new models; the sources give no specifications or release date.
- Anthropic's closed models still lead Arena.AI's text ranking, leaving the gap to Chinese systems unresolved.
- Pricing comparison for DeepSeek's V4-Flash comes from one research firm and is not independently confirmed.
China: State media present the open-sourcing by Alibaba and MiniMax as lowering costs for developers and users worldwide.
United States: US business media frame the launch as another Chinese AI breakthrough, while American labs OpenAI, Anthropic and Google keep their models closed-source.