Tencent released a preview of its next flagship large language model, Hy4, and published its weights as an open-source preview Hy4 preview.
The move follows Tencent’s recent in-house model push and is positioned as a capability upgrade meant to close gaps with rival Chinese and international labs, while shifting more product traffic onto the company’s own stack.
The Information reported the release and described it as significant progress in Tencent’s AI programme major progress.
Tencent also rolled the preview into multiple product surfaces as part of the launch launched across multiple products.
The open-source step mirrors a broader China strategy of shipping larger public weights to accelerate adoption and external evaluation, and it gives developers direct access to the model for experimentation.
Hy4’s debut builds on Tencent’s earlier Hy3 releases this year, which the company used to prototype stronger reasoning and agent behaviours inside its consumer and cloud services.
Analysts and competitors will now be able to benchmark the preview and weigh how quickly Tencent converts the preview into a full production release later in the year.
