Tencent's Hy3 Ranks Second Among Open-Weight Models in Frontend Code Arena
Tencent's Hy3 model has achieved the #2 rank among open-weight models and #16 overall on the Frontend Code Arena benchmark. It also placed in the top 20 for categories such as Reference-Based Design, Gaming, Simulations, and Content Creation Tools. This milestone highlights the rapid advancement of open-weight models in specialized tasks like frontend code generation, offering developers powerful, cost-effective alternatives to proprietary models. It demonstrates that open-source reasoning models are becoming highly competitive in practical software development workflows. While Hy3 secured the second spot for open-weight models, Kimi K3 currently holds the number one position on the Frontend Code Arena benchmark. Additionally, Hy3 has previously topped OpenRouter's global LLM leaderboard shortly after its open-source release.
## BACKGROUND
Frontend Code Arena is a benchmark designed to evaluate AI models on their ability to generate frontend and user interface code. Open-weight models are AI models whose neural network weights are publicly released, allowing developers to run, customize, and self-host them. Tencent's Hunyuan Hy3 is one such open-weight reasoning model that has gained traction for its speed and performance.