Tencent's Hy4 adds vision and computer use to its open model
The full Hy4 release replaces the text-only preview Tencent shipped in August, adding image understanding and the ability to drive a desktop by itself.
Tencent launched the full Hy4 model on 4 November, succeeding the preview it shipped in August. According to Tencent’s own announcement, this release introduces visual understanding and computer use “for the first time”, letting Hy4 operate both command-line and graphical interfaces rather than just generate text.
Tencent says Hy4’s hallucination rate fell from 12.95% in the preview to 7.2%, and that errors from unsupported assumptions dropped from 23.1% to 5.6%. In a blind evaluation by 213 financial professionals across 579 real-world tasks, Tencent reports Hy4 scored 3.04 against 2.97 for DeepSeek V4.1 Flash and 2.68 for both Kimi K3.1 Pro and GLM-5.3. Pricing stays the same as the preview: roughly $0.90 per million input tokens, $2.70 per million output tokens, and $0.045 per million cached input tokens.
Why it matters: open-weight models with working computer-use agents are still rare outside the big US labs. If the weights land as described, Hy4 becomes a self-hostable option for teams building desktop or browser agents, not just coding assistants.
The caveat: Tencent says the model weights and system card are coming “in the coming days”, not at launch, and the November announcement doesn’t restate the preview’s 770-billion-parameter, 1M-token spec sheet, so the architecture behind this release isn’t confirmed yet.