Tencent's Hy4 Can See Screens and Use a Computer — but You Can't Download It Yet
Photo: Hippopx (public domain). Illustrative image.
What happened. Tencent released Hy4 on 4 November 2026, the new flagship of its Hunyuan model line and the successor to the Hy4 preview from late August. It is a mixture-of-experts model with 770 billion total parameters, of which 49 billion activate for any given token — up from 295 billion total in Hy3, a roughly 2.6-fold increase in scale.
What is new. Two capabilities are firsts for the Hunyuan line. The first is visual understanding: Hy4 processes images at a fine-grained level, analyses complex charts, and reasons about mathematical visuals. The second is computer use: it can operate a computer through its graphical interface and command line, combining perception, deep reasoning and tool invocation — the same direction other frontier labs are pushing agents, but now from a major Chinese lab at flagship scale. Tencent also reports stronger instruction following, better long-document understanding, and longer-horizon agent task execution.
What you can actually use today — and what is only promised. The distinction matters, because much of the coverage frames Hy4 as an open-weight model. Available now, as a hosted service: Hy4 through Tencent Cloud TokenHub, WorkBuddy and CodeBuddy, Yuanbao, ima, and the WeCom assistant Dayuan, plus APIs on OpenRouter, Hermes and OpenCode. Not yet available: the open weights. Tencent says they will be open-sourced soon, but no date has been given, and until they are published nobody outside Tencent can download, run or fine-tune the model itself. Coming soon, also undated: Hy4 Flash, a lighter version focused on speed and cost, with no specifications or price yet.
Pricing, labelled by source. Tencent Cloud’s preview-level pricing is RMB 6 per million input tokens, RMB 18 per million output tokens, and RMB 0.3 per million cache-hit tokens. TokenHub carries a 50% discount until 31 December 2026, and WorkBuddy and CodeBuddy are free from 4 to 15 November 2026. OpenRouter lists $0.834 per million input tokens and $2.501 per million output tokens — a gateway listing price, not Tencent’s own price, and the two should not be quoted interchangeably. The context window is listed at 1,048,576 tokens on OpenRouter’s listing; treat that as a gateway figure rather than a vendor-confirmed specification.
The honest reading. Tencent’s own framing — labelled as Tencent’s, not as an independent finding — is that this is the first model in the Hunyuan line to combine visual understanding and computer use, and that its line-up is one of the strongest open-weight models available today. Tencent also admits, in its own materials, that Hy4 trails leading closed models on research maths. No independent benchmarks or hands-on results exist yet, and none are claimed here. The South China Morning Post reports that the Hy4 preview achieved a 23rd-place ranking on WebDev, a global AI benchmark leaderboard — a ranking for the preview, not for Hy4 itself.
Why it matters. Vision plus computer use in one flagship model from a major lab is a statement about where the frontier is heading: models that do not just generate text, but perceive screens and act on machines. The open-weights promise matters too — but the gap between “announced as open-weight” and “weights actually published” is where many launches quietly stall. What matters next is not the announcement, but the download, and whether the capabilities hold up when people outside Tencent finally get to test them.
Sources
- DEV Community — “Tencent’s Hy4 Can See and Use Your Computer — Explained Honestly” (4 November 2026): https://dev.to/paullovelace/tencents-hy4-can-see-and-use-your-computer-explained-honestly-4fc0
- SXF Brief — “AI News Nov 5, 2026: Tencent Hy4, Claude Biology, CrowdStrike” (5 November 2026): https://sxf.si/brief/
- Kimkj Daily Briefing — “AI, Economy, and Global Conflict Daily Briefing · Thursday, November 5, 2026”: https://kimkj.com/en/daily-briefing-en/?mod=document&uid=20887
