Headline:
Alibaba’s New Qwen Model Rivals Anthropic, DeepSeek
Summary:
Key Points
- Alibaba released Qwen3.8-Flash, a new open-weight AI model it says rivals top Anthropic and DeepSeek releases
- The 125-billion-parameter model uses a new architecture built for the coming Qwen 4 generation to cut costs
- The launch comes amid Alibaba's fast expansion of its AI model and tools portfolio this month
The Latest:
Alibaba released Qwen3.8-Flash, a 125-billion-parameter model with open weights that the company said performs competitively against Anthropic’s Opus 4.6 and DeepSeek’s V4-Flash, built on a new architecture designed for the upcoming Qwen 4 series to cut training and inference costs.
Details:
- Specs: Qwen3.8-Flash has 125 billion parameters, a measure of the model’s sophistication, Alibaba said.
- Performance: Alibaba said the model competes with Anthropic’s Opus 4.6 and DeepSeek’s V4-Flash, without giving detailed benchmark figures.
- Architecture: The release uses a design built for the coming Qwen 4 series that significantly cuts training and inference costs, per the company.
- Expansion: Alibaba also launched an international version of its enterprise AI agent tool, QwenWork, targeting Asia, Latin America and the Middle East.
- Background: Earlier this month, Alibaba upgraded its flagship Qwen3.8 Max model into its largest ever.
Between the Lines:
The release fits a broader pattern of Chinese firms shipping AI models that rival US performance on global benchmarks, per Bloomberg’s reporting.
What to Watch:
Alibaba hasn’t set a release date for the full Qwen 4 generation, and Qwen3.8-Flash’s real competitiveness will become clearer as developers use it more widely outside China.
:
