I expect Qwen to top Chinese models when 4 gets released, but a good showing here by Tencent Hunyuan team. Looks like they are not Benchmaxing.

What happened: A post noted that Hy4 by @TencentHunyuan is now ranked the #1 best Chinese model in Agent Arena, and the author expects Qwen to top Chinese models when version 4 is released. Key numbers: Over 12K+ real-world, long-horizon agentic sessions; Hy4 placed Tencent Hunyuan as the only Chinese lab in the top 5; +5.4% net. Why it matters: The author called it a good showing by the Tencent Hunyuan team and said they appear not to be Benchmaxing, while noting they release slowly and may fall far behind by their next model release.

A post noted that Hy4 by @TencentHunyuan is now ranked the #1 best Chinese model in Agent Arena, and the author expects Qwen to top Chinese models when version 4 is released.

The author called it a good showing by the Tencent Hunyuan team and said they appear not to be Benchmaxing, while noting they release slowly and may fall far behind by their next model release.