InjazAI

Tencent launched Hy3, an open-source (Apache 2.0) MoE model with 295B total parameters (21B active), outperforming larger models by 2-5x, available for free on OpenRouter until July 21, 2026, offering MENA teams a powerful and efficient model for production use.

1 min read

Tencent Launches Open-Source Hy3 Model with 295B Parameters, Outperforms Larger Models

Tencent Launches Hy3: An Open-Source Model with Competitive Performance

FAQ

What is Tencent's Hy3 model?

Hy3 is a Mixture-of-Experts (MoE) large language model with 295B total parameters and 21B active parameters, licensed under Apache 2.0, developed by Tencent.

How does Hy3 compare to other models?

Hy3 outperforms similar-sized open-source models and rivals models 2-5x larger in benchmarks, making it a cost-effective choice.

Is Hy3 available for use in MENA?

Yes, the model is free on OpenRouter until July 21, 2026, and can be downloaded from Hugging Face (598GB full, 300GB FP8).

What are ideal use cases for Hy3?

Hy3 suits production applications and creative tasks like generation and analysis, thanks to its efficiency and 256K context.

Source: Simon Willison (LLM & tools)

AI-assisted content, human-reviewed.