Editorial illustration for Tencent releases Hy4 AI with vision and computer use, claims second in China
Tencent Launches Hy4 AI Model With Vision and Computer Use
Tencent put out its new flagship AI model, Hy4, on Tuesday, two months after showing a preview version. The company says Hy4 ranks second among Chinese models, trailing only behind the field's current leaders, though it still lags well behind Claude Opus 5.5 on hard math and coding benchmarks. Unlike earlier Tencent models, Hy4 can analyze images, read fine detail in photos and diagrams, and operate software on its own, a capability known as computer use. Tencent built the model with coding, finance, and office tasks in mind, and says it's already put Hy4 to work improving training pipelines for its other AI systems.
For now, there's a catch. Tencent hasn't released the model weights, so Hy4 is only reachable through Tencent Cloud, the OpenRouter platform, or apps like the Yuanbao chatbot. That's a departure from Tencent's usual open-weight approach with past Hunyuan models, and the company hasn't said when, or if, it plans to open the model up. Tencent has pointed to one example to show off the image skills: spotting a roughly two-meter component on a warehouse roof buried inside a huge aerial photo.
An initial independent assessment from Artificial Analysis paints a slightly different picture. On its Intelligence Index, which combines knowledge work, coding, science, and general knowledge, Hy4 scores 35 points, above the median of 30. That ties it with Grok 4.7 and puts it in second place among Chinese lab models, behind Kimi K3.1 Pro at 46 points.
Why this matters
Tencent's framing here is the interesting part, not the benchmark position. A model that gets used to tune the training pipelines of its own successors is a different kind of release than a chatbot upgrade. If Hy4 and Hyra 2.0 genuinely improved how Tencent trains smaller models, that's a small step toward the self-improving loop labs have been talking about for years, and it's worth watching whether Tencent publishes any specifics on what actually changed in those pipelines.
We'd also flag the usual gap between claims and access: "second among Chinese models" is Tencent's own framing, weights aren't out yet, and "falls well behind" the leader is buried in the summary for a reason. For developers and founders building on Chinese open models, the real signal will come once Hy4's weights ship and outside groups can test the vision and computer-use claims directly. Until then, treat the coding, finance, and office-work optimization claims as marketing until someone outside Tencent runs the numbers.
Related Reading
Open Source Study: 6.7% of Deepfake Requests on Hugging Face Targeted Children
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
Open Source Axiom Math raises USD 200M Series A led by Menlo to keep AI slop out of math
From Archives OpenAI Releases Cybersecurity Model; Alpha in Unaudited Code via AI
From Archives AI Startup Potential Sciences Skips Lab Work in Drug Discovery
Common Questions Answered
What new capabilities does Tencent's Hy4 model have compared to earlier Tencent models?
Hy4 introduces three major new capabilities: image analysis, the ability to read fine detail in photos and diagrams, and computer use, which allows the model to operate software on its own. These features represent a significant expansion beyond the capabilities of Tencent's previous AI models.
How does Hy4 rank among Chinese AI models according to independent assessments?
According to Artificial Analysis's Intelligence Index, Hy4 scores 35 points and ranks second among Chinese lab models, tied with Grok 4.7 and behind Kimi K3.1 Pro which scores 46 points. However, Tencent claims Hy4 ranks second overall in China, though it still lags behind Claude Opus 5.5 on hard math and coding benchmarks.
What is significant about Tencent's approach to using Hy4 in their development pipeline?
The key significance is that Hy4 is being used to tune the training pipelines of its own successor models, representing a self-improving loop that labs have been pursuing for years. This approach differs from typical chatbot upgrades and suggests Tencent may be achieving incremental improvements in how it trains smaller models.
When was Tencent's Hy4 model officially released?
Tencent released Hy4 on Tuesday, approximately two months after showing a preview version of the model. This release marks the company's official launch of its new flagship AI model with expanded capabilities.
Further Reading
- Tencent releases Hy4 with vision and computer use, claims second place among Chinese AI models - The Decoder
- Tencent Launches Hy4 AI Model With Vision and Computer-Use Capabilities - TechRepublic
- Tencent launches Hy4 with vision and computer-use capabilities - TechNode
- Tencent Hy4: New AI model joins China’s top tier - Heise
- Tencent Declares Latest AI Model Ranks Among China’s Best - Bloomberg
