From the source
Tencent has released Hy4, a language model that understands images and operates software on its own, with weights promised at a later, unspecified date.
- The model was trained on significantly more data than its predecessor and, according to Tencent, ranks among the strongest open-weight models for the first time.
- Until the weights arrive, Hy4 is available only through Tencent Cloud, OpenRouter and Tencent apps such as the Yuanbao chatbot.
- Tencent trained it on professional tools including Blender and CAD programs, and worked with its own gaming studios to adapt it for Unreal, Unity and Godot.
- In a test with over 200 finance professionals on nearly 600 real-world tasks, Tencent says Hy4 ranked ahead of
Deepseek V4.1 Flash, Kimi K3.1 Pro and GLM-5.3 Flash, and a revamped search feature cut its hallucination rate from nearly 16 to about 5 percent in internal tests.
- Artificial Analysis instead scores Hy4 at 35 points on its Intelligence Index, second among Chinese lab models behind Kimi K3.1 Pro at 46, and on its AA-Omniscience test the model invents an answer 89 percent of the time when it does not know one.
- Tencent charges the same as for the preview, about $0.83 per million input tokens and $2.50 per million output tokens, and admits Hy4 still falls well behind the best commercial models on advanced math, longer coding projects and complex image analysis.