Tencent released Hy4 on 4 November 2026, the full version of its flagship Hunyuan model. For the first time in the series, the model understands images and can operate a computer — using the command line and graphical interfaces to carry out tasks, not just answer questions about them. It is live now through Tencent Cloud and OpenRouter, and inside Tencent's own apps Yuanbao, WorkBuddy and CodeBuddy, at unchanged API pricing of roughly $0.90 per million input tokens and $2.70 per million output tokens.

The vision and computer-use additions are the real story here. Earlier Hunyuan releases were text-only, so Hy4 is Tencent's catch-up move toward the agentic models its rivals have been shipping all year: a model that can look at a screen, understand what it shows, and then act on it. Tencent is also planning a cheaper Hy4 Flash variant, which suggests the company wants the model family to cover both heavy work and everyday, cost-sensitive use.

One honesty note matters: despite the "open" framing around the launch, the model weights are not yet released. Tencent says it plans to publish them in the coming days, along with a system card. Until that release actually lands, Hy4 is available as an API and inside Tencent's apps — not as a downloadable open model — so anyone planning to self-host or fine-tune it should wait for the weights before making plans.

Hy4 arrives in a week crowded with big model news, and its test will be the same as every flagship release: whether developers actually switch. The pricing is unchanged from the previous generation, the distribution is broad, and the computer-use capability is genuinely new for the Hunyuan line. If the promised open release follows quickly, Hy4 could become one of the most widely used Chinese models outside China; if it slips, the launch will read as an API update rather than the open-model moment Tencent is pitching.

Sources