Tencent has expanded international access to its latest artificial intelligence model, Tencent Hy3, as the company accelerates enterprise AI adoption with enhanced reasoning, agent capabilities and developer tools. Previously known as Tencent Hunyuan, Hy3 was officially launched on July 6 and is now available to global users, developers and businesses through multiple Tencent platforms.

The company said users can access Hy3 free of charge through WorkBuddy until August 31, 2026 (Pacific Time). The model is also integrated into Tencent Design Miora and Tencent Cloud TokenHub, with additional deployments planned across Tencent Cloud's AI-native product portfolio. Developers can integrate the model into applications using APIs, coding environments and third-party development tools.

Poshu Yeung, Senior Vice President of Tencent Cloud and Head of Tencent Cloud International, said the expanded rollout reflects Tencent's focus on helping organisations transition from AI experimentation to large-scale enterprise deployment.

"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," Yeung said. "By connecting advanced models with real-world scenarios, enterprise-grade engineering, flexible model orchestration and secure infrastructure, we aim to give enterprises the control and reliability they need to move from pilots to scaled deployments."

Tencent said Hy3 experienced rapid adoption following its launch, recording more than 68 times the API calls of its predecessor and topping OpenRouter's global large language model usage leaderboard within one week. The model is also available across developer platforms including Hermes, Kilo, Cline, OpenClaw, OpenCode and Cherry Studio, while remaining accessible through open-source communities such as Hugging Face and ModelScope.

Built on a hybrid fast- and slow-thinking Mixture-of-Experts (MoE) architecture, Hy3 features 295 billion total parameters with 21 billion active parameters and supports context lengths of up to 256,000 tokens. According to Tencent, the model delivers competitive performance in reasoning, coding, instruction following, in-context learning and AI agent tasks while using significantly fewer active parameters than many flagship AI models.

Tencent said Hy3 has been continuously refined using feedback generated from millions of users across its ecosystem. The company noted that this iterative development approach has improved the model's stability and expanded its applications across software development, office productivity, financial analysis, front-end design and game production while reducing token consumption for enterprise workloads.

The model powers WorkBuddy, Tencent's AI agent workspace that enables users to complete research, data analysis, document creation, presentations and workplace communications using natural language prompts. The platform supports concurrent multi-agent workflows with more than 100 built-in skills and integrates with collaboration tools including Discord, Slack and Telegram. Internal evaluations showed Hy3 achieved a task success rate exceeding 90 per cent while reducing average task completion time by 34 per cent compared with the previous Hy model.

Tencent has also integrated Hy3 into Tencent Design Miora, an AI-native creative platform for designers, marketers and content creators. The platform enables users to generate production-ready graphics, videos, 3D assets and user interface designs using natural language instructions while maintaining editable and traceable creative workflows.

For enterprise customers, Hy3 is available through Tencent Cloud TokenHub, a Model-as-a-Service platform that allows organisations to deploy and manage multiple large language models through a single API. The platform features intelligent model routing, usage governance, flexible billing and enterprise token management designed to simplify AI deployment at scale.

Tencent also announced that Hy3 has been released under the commercially permissive Apache 2.0 licence, allowing developers to modify, deploy and integrate the model into their own applications. API pricing on OpenRouter starts at US$0.1288 per million input tokens and US$0.5336 per million output tokens.

Yeung said Tencent Cloud International continues to work with enterprises across Asia Pacific, Europe, the Middle East and the Americas to integrate Hy3 into business environments through its broader AI infrastructure, including enterprise knowledge bases, agent development platforms and data management tools.

Tencent Hy3 Tencent AI Tencent Cloud enterprise AI AI model generative AI WorkBuddy Tencent Design Miora Tencent Cloud TokenHub AI agents large language model LLM Mixture of Experts MoE a