From the source
Tencent introduced Hy4, a frontier large language model that adds computer-use capabilities for the first time, using visual understanding, reasoning and tool-calling for long-horizon agentic work.
- The model follows Hy4 Preview, which arrived August 28 with 770 billion parameters and 49 billion active parameters; both offer a 1 million token context window, placing them alongside current open-weights frontier models from
Mistral AI, Moonshot and
Z.AI.
- Tencent says it scaled pre-training and post-training substantially, reporting gains over the preview in instruction following, long-context understanding, agentic long-horizon tasks and deep research, after identifying excessive reasoning and over-verification in the earlier version's thinking processes.
- The company says Hy4 improved 19.1 percent on the GameDev bench and 9.8 percent on GameLogic benchmarks, though the baseline numbers remain internal and were not released.
- Hy4 is available on Tencent Cloud TokenHub, WorkBuddy, CodeBuddy, Yuanbao, WeCom's AI Assistant and Ima, and through third-party sources including OpenRouter, Hermes and OpenCode.
- Tencent kept pricing unchanged from the preview at around $0.90 per million input tokens, $2.70 per million output tokens and $0.045 per million cached input tokens, and says it intends to release the model weights and system cards in the coming days.