Alibaba has previewed Qwen 3.8, a new AI model with 2.4 trillion parameters that the company claims ranks second only to Anthropic's Fable 5 in performance. The announcement, confirmed by Bloomberg, marks another step in China's push to close the gap in global AI competition.
Key Facts
- Qwen 3.8 features 2.4 trillion parameters and is released as an open-weight model
- Alibaba positions the model as the second-strongest solution after Fable 5
- The model competes against Kimi K3 and other local rivals
- Alibaba's stock surged 5 percent following the announcement
A Signal Against US Dominance
Releasing Qwen 3.8 as an open-weight model carries strategic weight: unlike proprietary models such as GPT-4 or Claude, developers can download, run locally, and fine-tune the model. This appeals especially to organizations seeking to reduce dependency on US vendors – a growing concern in China and the EU.
Qwen competes not only with frontier models like Fable 5 but also with other open-weight alternatives including DeepSeek, Mistral, and Gemma. According to available benchmarks, Qwen 3.8 can already run on commodity hardware with 24 GB VRAM – an advantage for developers with limited budgets.
Alibaba's Broader AI Strategy
The Qwen announcement coincides with reports that Apple has recruited members of the Qwen team to improve Siri in China. This suggests that Alibaba's AI expertise is recognized beyond China – a rare signal of quality in Chinese models.
The 5 percent stock surge following the announcement shows investors view the model as a strategic breakthrough. Analysts point to a $125 resistance level for Alibaba stock (NYSE: BABA), now in focus due to such developments.
What This Means for European Companies
For German AI practitioners and developers, Qwen 3.8 could become relevant – not as a replacement for frontier models like Fable 5, but as a cost-effective, locally deployable alternative for specialized tasks. Open-weight models reduce vendor lock-in and enable on-premises data processing. For regulated sectors (finance, healthcare, public administration), this is a genuine advantage.
However, questions remain about how Qwen 3.8 truly performs in European benchmarks and whether the company's performance claims have been independently validated. Those seriously evaluating the model should test it themselves – either on local hardware or via available APIs.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
