Tencent has released Hy4 preview, a new open-source AI model with 770 billion parameters, marking the Chinese tech giant's latest move in an increasingly competitive race to build powerful, freely available artificial intelligence.
The model, posted to Hugging Face on Friday, uses a mixture-of-experts (MoE) architecture, meaning that while it has 770 billion parameters in total, only about 49 billion are activated for any given text request. This design allows the model to be both powerful and computationally efficient, as it doesn't need to engage its full capacity for every task.
Hy4 is aimed at software engineering, research tasks, and financial analysis — areas where Tencent is betting that open-source AI can gain traction against proprietary models from OpenAI, Anthropic, and Google. The company plans to integrate Hy4 into its existing products, including CodeBuddy for developers and WorkBuddy for enterprise users.
Tencent acknowledged that the early-release model has limitations, noting it can sometimes take longer than necessary to work through complex questions and may over-verify its own answers. These kinds of candid disclosures are becoming more common as Chinese AI labs release models into the open-source ecosystem.
The release comes as Tencent ramps up its investment in artificial intelligence amid an intensifying domestic and international competition. The company unveiled its Hunyuan 3.0 large language model in April, its first major release since hiring former OpenAI researcher Yao Shunyu to lead its AI platform development.
Tencent's push into open-source AI mirrors a broader trend among Chinese tech companies. Alibaba has gained significant traction with its Qwen family of models, which have become among the most-downloaded open-source AI models globally. Baidu, ByteDance, and DeepSeek have also released competitive open-source offerings.
The open-source approach gives Chinese companies a strategic advantage in building ecosystems and developer loyalty, while also serving as a counterpoint to the more closed strategies favored by some US-based AI labs. By making powerful models freely available, companies like Tencent can drive adoption across industries and geographies without the friction of licensing barriers.
