Inside Tencent Cloud’s Agentic AI Push at SuperAI 2026

Key features include the ability to run multiple tasks and AI agents in parallel, with flexible multi-model integration via APIs . It is available as a Personal Edition desktop app and an Enterprise Edition that can be managed through a natural-language interface, including remote control from messaging platforms like Discord, Slack, and Telegram .

Under the hood, WorkBuddy runs on an upgraded Agent Runtime. Tencent reports a sandbox wake time of just 50 milliseconds—roughly 20 times faster than the industry baseline—which allows the system to release 70% of idle compute resources . This efficiency is critical for enterprise environments where cost and responsiveness are paramount.

Creative Automation: Tencent Design Miora

Miora targets a different user: designers and creative teams. It is an agentic creative studio that coordinates a network of specialist agents and skills to generate a complete, on-brand creative system from a single natural-language brief .

Miora’s output spans images, video, 3D assets, and user interfaces. Its standout feature is persistent memory, which allows the agent to retain brand guidelines, style preferences, and project context across multiple sessions, ensuring visual consistency across campaigns . Tencent is marketing the tool to both independent creators and enterprise teams in Southeast Asia, with the promise of collapsing weeks-long production cycles into hours .

One API for Many Models: TokenHub

TokenHub addresses the fragmentation of the large language model (LLM) market. It is a unified model gateway that aggregates Tencent’s own Hunyuan models alongside popular third-party models such as Kimi, GLM, and DeepSeek .

The platform covers a wide range of scenarios from dialogue and reasoning to code, vision, and image and video generation . For developers, TokenHub provides a single API to discover, compare, and invoke multiple LLMs without the operational overhead of managing separate infrastructure for each . The goal is to lower the barrier for enterprises to experiment with and deploy frontier models in production .

The Efficiency Engine: AI-Generated Code and Smarter Models

Tencent’s product launches are backed by internal efficiency figures that make a compelling case for its technology.

CodeBuddy, the company’s coding agent, is now used by over 90% of Tencent engineers, according to the company’s 2025 R&D Big Data Report. The result: 50% of all new code is generated with AI assistance, which has driven an overall R&D efficiency improvement of more than 20% . Average coding time has been cut by 40%, and in a particularly striking example, 90% of the code for Tencent’s own upgraded AI CLI tool was generated by the assistant itself .

On the model side, the new Hunyuan HY3 Preview demonstrates a similar focus on practical gains. The model is a 295-billion-parameter Mixture-of-Experts (MoE) architecture with 21 billion active parameters. Tencent reports a 40% improvement in inference efficiency compared to its predecessor, with input pricing as low as 1.2 yuan per million tokens . The company describes it as a leading reasoning and agent model for its size class . The efficiency gains are not just theoretical; in internal testing, Agent Memory systems improved pass rates on benchmark tasks while dramatically reducing token consumption .

A Strategic Pivot to Production

These announcements are not isolated product drops. They are the most visible manifestation of a broader strategic realignment for Tencent in 2026.

Tencent is aggressively moving from AI research and experimentation to large-scale, production-grade agent deployment. Chairman and CEO Ma Huateng stated that the year began with "significant initial progress on new AI products," and the company has been rapidly shipping a family of agents including CodeBuddy, WorkBuddy, and creative tools .

The scale of this internal transformation is further validated by an Omdia ranking of the World’s Most Practical Agentic-Driven Digital Service Groups, which placed Tencent fifth overall, making it the only China-based provider in the global top 5 . The company has framed AI as a "new engine" for R&D productivity, a claim substantiated by the 20% efficiency leap .

The choice of SuperAI 2026 as the launch venue was strategic. The conference, held at Marina Bay Sands on June 10–11, is Asia’s largest AI event, having sold out with 10,000 attendees from over 1,500 companies across 150 countries . Positioned as a "neutral intersection" for US, Chinese, European, and Asian AI development, it provided Tencent with a high-profile stage to court Southeast Asian enterprises . Tencent’s presence at Booth PB4, which also featured its PalmAI biometric technology, served as a physical demonstration of its “from experiment to deployment” narrative .