Tencent WorkBuddy Grows as Hy4 Agent Models Near Launch

Tencent WorkBuddy has expanded its enterprise agent push, with new capabilities unveiled for the office productivity agent and previewing the upcoming Hy4 model family at the 2026 Tencent Global Digital Ecosystem Summit in Shenzhen. The announcement, reported by Tencent on October 30, 2026, positions WorkBuddy as a central piece of the company's international enterprise artificial intelligence strategy, with the new models designed to handle longer and more complex workplace tasks.

WorkBuddy has grown quickly since its launch in March 2026. According to Tencent, the agent has handled more than nine billion task requests, with adoption spanning more than 50 industries. The company also reports that half of China's top 500 enterprises now use the product. Those figures frame the latest expansion as a scale-up of an already widely deployed tool rather than a first launch.

The timing matters for the broader agent market. Enterprise buyers are moving past experiments with single-purpose chatbots and toward agents that can complete multi-step work inside the tools employees already use. Tencent's pitch is that WorkBuddy meets teams where they work, connecting with widely used messaging applications and enterprise systems so tasks can be finished without switching contexts.

WorkBuddy brings agents into everyday enterprise workflows

Under the expansion, Tencent WorkBuddy is being offered more broadly to international enterprises. According to Tencent, the agent connects with messaging applications and enterprise systems, allowing teams to complete tasks within the tools they already use. Customers can choose the artificial intelligence model that best fits each task, while built-in security controls are intended to support enterprise use.

That model flexibility is a notable design choice. Rather than locking customers into a single model for every job, Tencent lets teams match the model to the task. The approach mirrors a wider industry shift toward model choice inside agent platforms, where routing, cost, and capability trade-offs differ sharply between simple lookups and long-running workflows.

In China, Tencent also introduced Buddy One, a cross-application collaboration system that lets WorkBuddy coordinate tasks across the company's office applications and specialized business agents. The idea is to reduce the handoffs that slow down office work: instead of one agent finishing a fragment and handing it to another tool, the system coordinates across applications as a single workflow.

Tencent has also launched a wider international agent portfolio alongside WorkBuddy. According to the company, that portfolio includes CodeBuddy for coding work, Miora, the Agent Development Platform, MAGIC AI, and Agent Runtime. Together, the products are meant to let Chinese and international teams collaborate on the same artificial intelligence workbench, a pitch aimed at multinational companies with staff on both sides of that divide.

Hy4 and Hy4 Flash aim at long-horizon agent work

The other half of the announcement is the Hy4 model family. Tencent said the general release of Hy4 is coming soon, with stronger performance on long-horizon and real-world productivity tasks, improved reasoning efficiency, and visual understanding for the first time. A second model, Hy4 Flash, is being prepared as a cost-effective option that carries over the visual understanding and agentic capabilities of Hy4.

Long-horizon performance is the capability enterprise agent builders watch most closely. Short tasks are largely solved; the hard problems appear when an agent must plan across many steps, recover from errors, use tools in the right order, and keep context intact over hours of work. Tencent's emphasis on reasoning efficiency suggests the company is targeting that gap directly, aiming for agents that can sustain complex work without ballooning cost or losing track of the goal.

Visual understanding adds another dimension. Many workplace tasks involve documents, screenshots, charts, and interfaces that are not pure text. A model that can interpret visual input alongside text can handle a wider range of productivity jobs, from reviewing a slide deck to working through a form or dashboard. Hy4 Flash is positioned to bring that capability to cost-sensitive deployments, which could matter for high-volume use cases where the flagship model would be too expensive to run at scale.

Enterprises can access Hy models alongside leading third-party models through Tencent Cloud TokenHub, a unified platform with a single application programming interface. That single-API access matters for developers, who can experiment with different models without rebuilding integrations for each one. It also signals that Tencent expects its models to compete on merit inside mixed-model stacks rather than only inside a closed ecosystem.

What the WorkBuddy expansion means for agent builders

For agent builders, the Tencent WorkBuddy expansion is a reminder that the enterprise agent race is increasingly about distribution and workflow integration, not model benchmarks alone. A capable model still needs to reach workers inside their existing tools, respect enterprise security controls, and coordinate across applications to be useful. Tencent's combination of a deployed productivity agent, a cross-application coordination layer, and a new model family is designed to cover all three at once.

The move also sharpens competition with other enterprise agent platforms. Recent weeks have seen major vendors open their agent harnesses and developer tooling to outside builders, as covered in reporting on Microsoft opening its Copilot agent harness through Foundry. At the same time, vendors are adding supervision and governance layers for agents that touch sensitive operations, a trend reflected in Persona bringing supervised agents to identity operations. Tencent's emphasis on model choice and built-in security controls fits the same pattern: enterprises want agents that are powerful, but also governable.

There are open questions worth watching. Tencent has previewed Hy4 rather than releasing it, so independent benchmarks and real-world deployment data are still to come. The nine billion task requests figure shows scale, but it does not by itself show how many of those tasks were completed without human correction, or how the agent performs on the hardest long-horizon jobs. Those answers will come from customer deployments and third-party evaluations after the general release.

Still, the direction is clear. WorkBuddy has moved from a domestic productivity launch to an international enterprise agent platform in about seven months, and Hy4 is designed to give it stronger reasoning, visual understanding, and efficiency for the next phase. For teams building or buying enterprise agents, Tencent's announcement is another sign that the agent workplace is consolidating around platforms that combine models, tools, and workflow coordination in one place.

Sources: reporting on the announcement from Tencent, with additional coverage from TechBeat Philippines and TechNode Global.