---
type: "post"
id: "tencent-launches-hy4-joining-the-ranks-of-the-4c3756"
slug: "tencent-launches-hy4-joining-the-ranks-of-the-4c3756"
title: "Tencent Launches Hy4, Joining the Ranks of the World's Leading Open-Source AI Models"
description: "Tencent announced Hy4, a new-generation language model that adds visual understanding and computer use to the Hy line."
retrieval_nugget: "Tencent announced Hy4, a new-generation language model that adds visual understanding and computer use to the Hy line."
published_at: "2026-11-05"
updated_at: "2026-11-05"
record_date: "2026-11-05"
date_kind: "published_at"
topics: ["computer-use"]
entities: ["Tencent","Hy4","Tencent WorkBuddy","WeCom","Tencent Financial Technology","Rabyte","DeepSeek-V4.1-Flash-1017","Kimi K3.1 Pro"]
related_patterns: ["agent-ready-software-exposes-capabilities"]
novelty_label: "new_fact"
what_changed: "The story adds Hy4’s ability to operate a computer, work in Blender and CAD, and analyze charts."
prior_refs: []
source_url: "https://www.tencent.com/tencent-launches-hy4-joining-the-ranks-of-the-worlds-leading-open-source-ai-models"
source_urls: ["https://www.tencent.com/tencent-launches-hy4-joining-the-ranks-of-the-worlds-leading-open-source-ai-models"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:tencent-launches-hy4-joining-the-ranks-of-the-4c3756"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/tencent-launches-hy4-joining-the-ranks-of-the-4c3756/","markdown":"https://newruntime.com/posts/tencent-launches-hy4-joining-the-ranks-of-the-4c3756.md","json":"https://newruntime.com/posts/tencent-launches-hy4-joining-the-ranks-of-the-4c3756.json"}
markdown_url: "https://newruntime.com/posts/tencent-launches-hy4-joining-the-ranks-of-the-4c3756.md"
---

# Tencent Launches Hy4, Joining the Ranks of the World's Leading Open-Source AI Models

## Retrieval answer

Tencent announced Hy4, a new-generation language model that adds visual understanding and computer use to the Hy line.

## Summary
Tencent announced Hy4, a new-generation language model that adds visual understanding and computer use to the Hy line. It can work through command-line and graphical interfaces, call tools, and handle tasks Tencent says were shaped by professional software workflows. For builders, the announcement is notable for its breadth, but the announced evaluations and forthcoming weights still leave important questions about reproducibility and access.

## What happened
Tencent announced Hy4 on November 4, 2026, describing it as its latest-generation large language model and a significant performance and capability step. The company says it plans to release model weights and system cards in the coming days, with cybersecurity and biological safety among its stated open-source development priorities. The announcement does not provide a specific license or a distribution location in the supplied material.

Hy4 introduces visual understanding in the Hy line, alongside improvements to instruction following, long-context comprehension and long-horizon agentic tasks. Tencent says the model supports fine-grained image processing, complex chart analysis and mathematical visual reasoning through visual understanding, reasoning and tool-calling capabilities. It also adds computer-use capabilities for the first time: Tencent describes interaction with both command-line interfaces and graphical user interfaces. The company says it has specially trained the model on professional software, including Blender and CAD, to improve planning and execution on sophisticated, extended tasks.

The announcement emphasizes Tencent WorkBuddy, where Hy4 achieved an average benchmark score of 83.18 and a 96.15 percent pass rate. Tencent says it ranked first in nine of 12 scenarios within the mainland Chinese market. The company describes applications including visual creative work, 3D scene construction, web development and document presentation. It also says WorkBuddy, when integrated with WeCom, can connect workflow data across chats, emails, documents, calendars and meetings, then track project updates, identify outstanding tasks and flag preparation needs and approaching deadlines. A “Projects” feature with new team collaboration capabilities is described as upcoming, not already released.

Tencent also reports a blind financial-analysis evaluation conducted with Tencent Financial Technology. It involved 213 financial professionals, including 153 experts from external institutions; about 70 percent had more than five years of industry experience. Across 579 real-world financial tasks, Hy4 scored 3.04, compared with 2.97 for DeepSeek-V4.1-Flash-1017 and 2.68 each for Kimi K3.1 Pro and GLM-5.3. The company further cites a 12.03 percentage-point improvement over Hy4 Preview on an internal Investment Research Agent Benchmark evaluation conducted by financial AI company Rabyte. In internal evaluations, Tencent says targeted search optimization lowered the hallucination rate from 12.95 percent for Hy4 Preview to 7.2 percent, and errors attributed to unsupported assumptions from 23.1 percent to 5.6 percent.

For development, Tencent says Hy4 is optimized for complex business logic, state management and front-end/back-end orchestration. It reports collaboration with Tencent Interactive Entertainment Group developers and integration work involving Unreal Engine, Unity and Godot. Compared with Hy4 Preview, the company reports improvements of 19.1 percent on GameDev and 9.8 percent on GameLogic benchmarks. The supplied announcement excerpt does not include further details on the scientific-research capabilities it begins to describe.

## What the archive adds
We have not covered this before; it opens a new thread. The archive does contain earlier Telegram posts about Kimi K3’s frontend-code evaluation and Tencent’s CubeSandbox runtime, but neither establishes prior coverage of Hy4 or this release.

## Trend signal
Hy4’s announced ability to use command-line and graphical interfaces, call tools and carry out longer tasks is relevant to [agent-ready software exposes capabilities](https://newruntime.com/patterns/agent-ready-software-exposes-capabilities/). The announcement describes model capabilities, not a specific permissioned interface exposed by an application, so it is an early and limited signal rather than proof that the pattern is established.

## Why now
Tencent dates the announcement November 4, 2026, and says weights and system cards are expected in the coming days. That near-term release could make the announcement’s claims easier for outside builders to inspect, subject to the eventual artifacts and their terms.

## Why it matters
For teams building agents, computer interaction and tool use matter because they connect model reasoning to work performed in software rather than stopping at generated text. Hy4’s announced coverage of visual inputs, command-line and graphical interfaces, and professional applications points toward a model intended to operate across different work surfaces. Its WorkBuddy examples also illustrate a product strategy in which the model is paired with workflow data and task-specific capabilities.

The practical value is not established by the announcement alone. The reported WorkBuddy and financial results are company-reported evaluations, and they do not tell a prospective user how Hy4 will perform on that user’s own tasks, what it costs to run, or how safely it behaves when given access to real systems. Builders should treat the reported results as reasons to evaluate the model when weights become available, not as a substitute for testing permissions, reliability and task completion in their own environment.

## Limits and uncertainty
The announcement is a primary source for what Tencent says it has built, but most performance figures are company-reported, and the described tests are not independently verified in the supplied material. It does not provide benchmark protocols, sample prompts, uncertainty ranges, or enough detail to compare results across evaluation settings. The promised weights and system cards are described as forthcoming; their exact release date, license, hosting location and usage conditions are unspecified. The material also does not establish API availability, inference pricing, hardware requirements, computer-use safeguards, or how the model performs outside Tencent’s products and selected evaluations.

## Reader takeaway
Hy4 is an announced model release with an unusually broad set of stated capabilities, including visual reasoning and computer use. Tencent offers useful evaluation numbers, but they remain the company’s claims until outsiders can inspect the methods and test the model. Watch for the promised weights and system cards before deciding whether Hy4 is ready for a production workflow.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Computer Use](https://newruntime.com/topics/computer-use.md): Explore the computer-use topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)