Tencent is opening its Hy3 large language model to global users, extending the model beyond its home market through the WorkBuddy agent workspace, the Miora design studio, and Tencent Cloud’s TokenHub platform, the company announced on August 5, 2026. The rollout broadens international access to the model Tencent’s Hunyuan team — now rebranded Tencent Hy — officially released on July 6, 2026.
Under the expansion, Hy3 is available free of charge on WorkBuddy to users worldwide until August 31, 2026 (Pacific Time), and through Tencent Design Miora and Tencent Cloud TokenHub. Developers can connect via API and integrate the model with coding extensions, third-party platforms, and other tools.
Hy3 is a hybrid fast-and-slow-thinking model built on a Mixture-of-Experts architecture, with 295 billion total parameters and 21 billion active parameters, supporting a context length of up to 256K. Tencent says it performs comparably to flagship models with two to five times as many parameters across reasoning, instruction understanding, in-context learning, code generation, and agent capabilities. The weights are available under the commercially permissive Apache 2.0 license on Hugging Face and ModelScope.
“Hy3 marks a significant step in how Tencent Cloud International is helping organisations move beyond AI experimentation and put intelligence to work in practical, trusted and scalable ways,” said Poshu Yeung, Senior Vice President of Tencent Cloud and Head of Tencent Cloud International, in the announcement.
What Hy3 is and where it came from
Hy3 is the output of a rebuild of Tencent’s model program that the company says began in late January 2026. The first model in the generation, Hy3 preview, launched on April 23, 2026, improving on the prior Hy2 in complex reasoning, instruction-following, in-context learning, code generation, and agent capabilities. The July 6, 2026 official release built on the preview through strengthened reinforcement learning and improved data quality and diversity.
Since the preview’s launch, Tencent reports that average daily token consumption rose twenty-fold, and that the number of WorkBuddy users actively selecting Hy3 preview grew sixfold. After the July release, Hy3 recorded more than 68 times as many API calls as its previous-generation model and, the company says, ranked first on OpenRouter’s global LLM usage leaderboard within one week of launch.
The model is also integrated across Tencent’s own product surface, including the Yuanbao assistant, the ima knowledge-base tool, the Marvis OS-level agent, and customer-service assistants on Weixin Official Accounts. On pricing, Hy3 API access on OpenRouter is listed starting at US$0.1288 per million input tokens and US$0.5336 per million output tokens.
How the claims are sourced
The capability figures here are Tencent’s own. The “two to five times” parameter-efficiency comparison, the internal WorkBuddy evaluation reporting a task success rate above 90% and a 34% reduction in average task completion time versus the prior Hy iteration, and the OpenRouter ranking are all vendor-reported, drawn from the company’s announcement and internal evaluations rather than independent testing. The 295B/21B architecture and 256K context are stated in the model documentation. The 90% task success rate and 34% completion-time reduction come from internal evaluations conducted within Tencent’s workplace environment, per the announcement.
What happens next
Access on WorkBuddy is free until August 31, 2026 (Pacific Time), after which the free tier ends. Tencent says further Hy3 integrations are planned across Tencent Cloud’s AI-native product portfolio, and regional partnerships with South Korea-based Cafe24 and Japan-based Metelix will distribute the model through those companies’ AI gateway and platform services.
