Domestic AI companies, including DeepSeek, have demonstrated that open-source strategies can carve out unique market niches. In response, a major Chinese internet company has announced that its flagship model will also be open-source for the first time.
Yesterday, the company unveiled its latest generation base model, Qwen3.8-Max, which contains 2.4 trillion parameters. Generally, the more parameters a model has, the better it can recognize patterns, produce accurate responses, and perform complex tasks.
The company claims that this model ranks among the top large language models globally, excelling in programming, scientific research, and long-term projects.
What sets this model apart is a shift in the company’s overall strategy. It plans to release the model’s weights next week, making it open-source.
Since introducing the Qwen series earlier this year, the company has released multiple open-source models. However, its Max series was previously closed-source, accessible only through APIs without allowing users to download the full model weights.
This change in approach follows the success of other Chinese companies that proved open-source and low-cost strategies can effectively grow market share. Rather than relying solely on profit from closed-source APIs, the company is now adopting an open-source approach to strengthen its influence in the evolving AI assistant ecosystem.
Some experts note that going open-source may slow down commercialization, but it also reduces barriers to adoption, broadens international reach, and encourages deployment across various cloud platforms and hardware types. Additionally, competitive open-source models can challenge high API pricing models from closed-source providers.
Qwen 3.8-Max’s estimated cost is around $6 per million output tokens, compared to about $50 for Anthropic’s Claude Fable 5. In AI workflows that involve frequent tool use, extended code generation, and iterative testing, this cost difference can significantly impact the commercial feasibility of the models.
However, smaller startups like DeepSeek face challenges related to computing resources. The open-source community has reported capacity issues due to an unprecedented surge in traffic, with some users experiencing errors, and urgent fixes are in progress.