Tencent has rolled out its 295-billion-parameter large language model, Hy3, making it accessible to users and developers worldwide rather than limiting it to the Chinese market. Standing out with its hybrid Mixture-of-Experts (MoE) architecture, the model aims to deliver an ecosystem of AI agents capable of automating complex workflows such as research, software development, and data analysis.
Technical Specifications and Architectural Design
Challenging its global rivals with its technical infrastructure, Hy3 boasts a total of 295 billion parameters, of which 21 billion are active at any given time. This hybrid MoE approach combines high processing performance with lower operational costs.
Key technical highlights of the model include:
- Context Length: Support capacity of up to 256,000 tokens.
- Performance Domains: Reasoning, code generation, financial modeling, and game development.
- Efficiency: According to Tencent's internal testing, the model reduces task completion times by 34% compared to previous models and delivers a success rate exceeding 90%.
Access Channels and Developer Support
To help developers and end-users adapt to the model, Tencent has provided various platform integrations and cost advantages. Hy3 can be accessed via WorkBuddy, Tencent Design Miora, and Tencent Cloud TokenHub.
Furthermore, the model is available for commercial use under the Apache 2.0 license, granting developers significant flexibility. According to OpenRouter data, it is offered at competitive pricing rates of $0.1288 per 1 million input tokens and $0.5336 per 1 million output tokens, while the WorkBuddy service remains free for global users until August 31, 2026.
Industry Implications
Tencent's decision to launch Hy3 globally with cost-effective API options is poised to intensify competition in the enterprise AI market. Thanks particularly to its open-source license and hybrid MoE architecture, companies are now better positioned to build their own custom AI agents at lower operational costs.
Frequently Asked Questions
What costs will developers face if they want to integrate the Hy3 model into their own applications?
Since the model is available under an Apache 2.0 open-source license, developers who wish to do so can run it on their own hardware for free. For cloud-based API usage, pricing through OpenRouter starts at $0.1288 per 1 million input tokens and $0.5336 per 1 million output tokens.
What exact role do Hy3's AI agents play in business processes?
Operating through the WorkBuddy platform, these agents can execute multiple tasks simultaneously via natural language commands to conduct research, perform data analysis, draft reports, and autonomously manage complex workflows without requiring human intervention.
*This news report is based on data originally published by Webtekno — Artificial Intelligence.
💬 Comments
No comments yet. Be the first!
You must be logged in to comment.
🔑 Log In