DatafetchPro

AI Infrastructure Tools

Tencent Hy

Tencent's open MoE model — 80B total, 13B active, fast and efficient

Votes
313
Followers
312
Comments
1
Best daily rank
#15

About Tencent Hy

Tencent Hunyuan-A13B

Open-source Mixture-of-Experts LLM with 80B total parameters and 13B active parameters.

  • Hybrid thinking: switches between fast and slow reasoning for cost-efficient inference
  • Apache-2.0 licensed and available on GitHub, Hugging Face, and Tencent Cloud
  • Fine-grained MoE architecture built for code, long-form text, and agent execution
  • Production ready: Docker images, vLLM support, FP8 and GPTQ-Int4 variants

Built by Tencent Hunyuan team as part of Tencent Hy family. Delivers frontier quality without running the full 295B Hy3 model, ideal for teams needing open weights with commercial-friendly license.

Launch history1

  1. Tencent Hy4

    Featured

    Tencent’s flagship model learns to see and use a computer

    See the whole day — 5 November 2026
    80votes