# Pandaily — Tencent’s Hy4 Adds Vision and Computer Use, Joining the Top Tier of Open-Weight Models

- Company: Pandaily (pandaily.com)
- Announced: 2026-11-04T09:22:06+00:00
- Category: open-weight-release
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://pandaily.com/tencent-hy4-vision-computer-use-open-weight-model-workbuddy
- Record: https://forck.live/items/31461-tencent-s-hy4-adds-vision-and-computer-use-joining-the-top-tier-of-open-weight
- Subject: Reporting
- Open weights: yes
- Models affected: Hy4, Hy4 preview

Tencent released Hy4, the latest generation of its large language models, which the company says ranks among the world's most capable open-weight models for the first time. The model adds vision capabilities and Computer Use for the first time, working across command line and graphical interfaces, with targeted training on software such as Blender and CAD applications. Tencent says Hy4 improves instruction following, long-context understanding and long-horizon agentic capabilities over Hy4 preview, and that it will release the model weights shortly along with a system card. A WorkBuddy-focused evaluation put Hy4's average quality score at 83.18 and its pass rate at 96.15%, and in a blind evaluation by 213 finance experts it ranked first among the open-weight models tested across 579 real-world finance tasks. Compared with Hy4 preview, Hy4 improved by 19.1% and 9.8% on the GameDev and GameLogic evaluations. On safety, Tencent says Hy4 was trained only on defensive cybersecurity data and that audits found no notable offensive risk, while its open-ended experiment design still lags frontier models by a wide margin. Hy4 is now live across Tencent's product portfolio, and its API is available through third-party platforms including OpenRouter, Hermes and OpenCode.

## Evidence

Verbatim from https://pandaily.com/tencent-hy4-vision-computer-use-open-weight-model-workbuddy:

> On November 4, Tencent released Hy4, the latest generation of its large language models, marking an across-the-board capability leap over Hy4 preview. With this release, Hy4 ranks among the world's most capable open-weight models for the first time.
> 
> Tencent will release Hy4's model weights shortly and publish a system card for the model.
> 
> Multimodal Understanding and Computer Use
> 
> Hy4 significantly expands both data and training scale across pre- and post-training, and introduces vision capabilities for the first time, delivering comprehensive improvements in instruction following, long-context understanding and long-horizon agentic capabilities.
> 
> Its new vision capabilities combine visual understanding, reasoning and tool use, enabling Hy4 to interpret dense charts, parse documents and analyze information across multiple images. The model can also combine visual information for multi-step analysis and deduction, moving beyond basic visual understanding to visual reasoning.
> 
> Building on its visual understanding, Hy4 now supports Computer Use for the first time, working across both the command line and graphical interfaces to handle a wide range of work and everyday tasks.

---

Record: https://forck.live/items/31461-tencent-s-hy4-adds-vision-and-computer-use-joining-the-top-tier-of-open-weight
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md