Main article
Visual understanding widens the tasks Hy4 can attempt
Tencent says Hy4 introduces the ability to understand images within the Hy model line, including complex chart analysis and mathematical questions presented visually. That expands the material a workplace assistant can consider beyond text alone. A chart, for example, contains relationships in its layout as well as words and numbers. Reading those relationships may help an assistant interpret a document, although the claim does not guarantee that every axis or visual detail will be understood correctly.
The release follows Hy4 preview, introduced in late August. Tencent describes broader training and improvements in agentic capabilities, coding and workplace automation. Here, agentic capability means the ability to plan and pursue tasks using tools across multiple steps. It should be judged by completed work and recoverable mistakes, rather than by fluent descriptions of a plan. Visual understanding can help with such tasks, but recognising information and acting on it are different stages.
The distinction matters in ordinary office work. An assistant may interpret a diagram correctly but still use the wrong document version when preparing a summary. Conversely, a successful tool call does not prove that the underlying interpretation was sound. These are examples of evaluation questions, not reported failures of Hy4. They explain why the model’s wider capabilities need assessment within complete workflows, where input, reasoning and final output can all affect the result.
















