Submit Tool
AI News

Tencent Open‑Sources 295B MoE Model Under Apache 2.0

Tencent releases its 295‑billion‑parameter MoE model Hy3 under Apache 2.0, removing limits and offering free OpenRouter access for two weeks now.

Suman Rana
Suman Rana
2 days ago
1 min read Updated Open Source
295B MoE AI model released under Apache 2.0 license
AI News · July 2026
Photo: Trend Tracker

Key Takeaways

The Hunyuan team at Tencent has launched Hy3, a 295‑billion‑parameter Mixture‑of‑Experts model with 21 billion active parameters.

The model follows a preview release and is now made available under the permissive Apache 2.0 license, removing earlier limits that barred use in the EU, the UK and South Korea.

This shift aligns with a broader industry trend of opening large AI models to accelerate research and commercial adoption, and the Apache 2.0 terms allow commercial use, modification, and distribution without royalties.

Previously, licensing restrictions prevented many enterprises from deploying the model, especially those operating in restricted regions.

Tencent is also offering the model free on OpenRouter for a two‑week period to encourage experimentation.

Community reactions on X have highlighted the license change as the headline development, suggesting Tencent could become a leading player in open‑source AI.

Key points:

  • 295 billion parameters, 21 billion active.
  • Apache 2.0 license replaces earlier preview terms.
  • Removes regional deployment blocks.
  • Free access on OpenRouter for two weeks.
  • Community sees Tencent as an emerging open‑source leader.

The release signals a commitment to a more open ecosystem and may influence other companies to reconsider their licensing approaches.

Potential Impact Areas

Potential impacts include:

  • Enterprises in the EU, UK and South Korea can now legally deploy the model without navigating complex licensing hurdles.
  • Developers can experiment with a 295 billion‑parameter MoE architecture, accelerating innovation and reducing reliance on closed APIs.
  • Startups gain access to a powerful model at no cost for a limited time, lowering barriers to prototype AI services.
  • Industry competition may increase as other companies consider permissive licensing to attract users.
  • Open‑source communities may see faster model fine‑tuning and benchmarking, raising overall standards.

Our Insight

Our insight is that Tencent’s move could reshape how large models are released and adopted.

The shift to Apache 2.0 removes legal barriers for many regions, which may encourage other vendors to adopt similarly open licenses.

For developers, the free two‑week access on OpenRouter offers a chance to benchmark the model against rivals such as Meta’s Llama series.

However, the model’s size and MoE design still require substantial computational resources, limiting immediate commercial rollout for smaller firms.

Risks include potential quality gaps compared to proprietary counterparts and the short window of free access, which may create pressure to monetize quickly.

Overall, the release signals a notable step toward broader open‑source AI, but its long‑term impact will depend on how the community utilizes the model and whether Tencent follows through with sustained support.

External Credit

Original source: venturebeat.com

Full credit goes to the original publisher. We link to this content for informational and commentary purposes only.

Disclaimer

This article is a curated summary and analysis. All credit goes to the original source. We aim to provide context and insights for the AI community.
Share:
Suman Rana
Article Author
Suman Rana
Menu
Home AI Tools Prompts Repos Contact Us About Us Privacy Policy Terms and Conditions
Submit Tool

Get the Daily Digest

AI trends, tools, and stories every morning. Free forever.