Archive · 27 Aug 2026
Alibaba Releases Smaller, Lower-Cost Qwen Model for Global Adoption
Alibaba has made a new AI model called Qwen3.8-Flash available for people to download.
It is designed to cost less to train and operate than some larger AI systems.
Alibaba says the model works about as well as several new models from other companies.
The model has 125 billion parameters, which are settings that help it make decisions.
Alibaba also shared the model’s weights, allowing others to use and study them.
The company hopes the lower cost will help more people and businesses adopt Qwen.
Alibaba is investing heavily in AI and has raised money to support that effort.
It also released an international version of its QwenWork AI tool.
Alibaba released its Qwen3.8-Flash AI model for download on Wednesday.
The model has 125 billion parameters and uses publicly revealed weights.
Alibaba says Qwen3.8-Flash performs competitively with Anthropic’s Opus 4.6 and DeepSeek’s V4-Flash.
The model uses architecture designed to reduce training and inference costs for future Qwen4 models.
Alibaba also released QwenWork internationally and raised HK$80 billion to support its AI expansion.
- Who
- Alibaba Group released the Qwen3.8-Flash model.
- What
- The company introduced a downloadable, lower-cost AI model with 125 billion parameters and publicly revealed weights.
- When
- The release took place on Wednesday.
- Where
- The model was released for global use, while QwenWork is targeting Asia, Latin America, and the Middle East.
- Why
- Alibaba aims to reduce AI training and inference costs, expand adoption, and strengthen its position in global AI.
- Model
- Qwen3.8-Flash
- Parameters
- 125 billion
- Availability
- Downloadable, with weights revealed
- Performance comparison
- Alibaba says it is competitive with Anthropic’s Opus 4.6 and DeepSeek’s V4-Flash
- AI investment commitment
- More than 380 billion yuan over three years
- Recent fundraising
- HK$80 billion raised through a Hong Kong follow-on share offering
- Additional release
- An international version of the enterprise-focused QwenWork AI agent