Tencent launches Hy4 with first-time vision and computer use
On 4 Nov 2026 Tencent announced Hy4, its latest-generation large language model, saying it now ranks among leading open-source models. Hy4 adds vision understanding and computer-use for the first time, keeps preview-era API pricing, and plans to release model weights and system cards in the coming days.
Verified against Tencent’s company news page (4 Nov 2026) and the official @TencentHunyuan post (2026-11-04T06:43:10Z): Tencent announced Hy4, its latest-generation large language model, and said it now ranks among the world’s leading open-source AI models. Tencent Hy also said it plans to release model weights and system cards in the coming days.
What changed (per the official release)
- Vision understanding for the first time, including fine-grained image processing, complex chart analysis, and mathematical visual reasoning.
- Computer-use for the first time, interacting with both command-line and graphical interfaces; the company says Hy4 received specialized training on professional software such as Blender and CAD to improve long-horizon planning and execution.
- The release also cites gains in instruction following, long-context comprehension, and long-horizon agentic tasks.
Availability and pricing
- Available across Tencent Cloud TokenHub, WorkBuddy, CodeBuddy, Yuanbao, WeCom’s AI Assistant (Dayuan), ima, and more; API access is also offered via third-party platforms such as OpenRouter, Hermes, and OpenCode.
- Despite performance gains over Hy4 Preview, Tencent says pricing is unchanged: about US$0.90 per million input tokens, US$2.70 per million output tokens, and US$0.045 per million cached input tokens.
- Hy4 Flash is scheduled for release soon for higher cost-efficiency scenarios.
- WorkBuddy and CodeBuddy are free from 4–15 Nov 2026; Hy4 on TokenHub is 50% off through 31 Dec 2026.
Safety statements (vendor)
- Cybersecurity: training prioritizes defensive data; offensive exploitation materials are excluded. Assessments by Tencent’s Xuanwu Lab and Keen Security Lab found no significant concerns about offensive cyber capabilities in the areas evaluated, per Tencent.
- Biosafety: strong foundational biological knowledge, but more limited open-ended experimental design than frontier models, per Tencent.
Vendor-reported evaluations (not independent verification) Tencent published internal/partner scores, including a financial blind evaluation where Hy4 scored 3.04 versus DeepSeek-V4.1-Flash-1017 at 2.97 and Kimi K3.1 Pro / GLM-5.3 at 2.68 each. Treat these as company-reported figures, not third-party verification.
This item was held earlier when only media teasers were available; it is now pushed after the official company page and account post were verified.