Tencent Hunyuan Hy4 to Be Released Soon: Preview Version Ranks 10th Globally on Agent Arena, Task Cost Lower Than Kimi…
Arena.ai revealed that Tencent Hunyuan's next-generation model Hy4 is about to be released. Its preview version ranked second among open models and 10th overall globally in Agent Arena's evaluation of more than 14,500 real agent task sessions, second only to Kimi K3 (Max). The Hy4 preview version had a net improvement rate of 5.4%, compared with 6.6% for Kimi K3 (Max); its median cost per task was about $0.26, roughly 68% lower than Kimi K3 (Max)'s $0.80, and it entered the Pareto frontier of that leaderboard. Tencent Hunyuan has not yet announced the release date, model parameters, or full technical details of the official version.
Why read this
AI score 59
Arena.ai's evaluation data provides the ranking and per-task cost of the Hy4 preview version on agent tasks, allowing a comparison with Kimi K3 (Max) to see the cost-effectiveness differences among open models.
Topics
Body · AI translation
Tencent Hunyuan Hy4 to Be Released Soon: Preview Version Enters Global Agent Leaderboard Top 10, Task Cost 68% Lower Than Kimi
Phoenix Tech Download the App for Exclusive Early Access November 04, 2026 07:13:39 From Beijing
Phoenix Tech News, November 4 (Beijing Time) — AI model evaluation platform Arena.ai recently revealed that Tencent Hunyuan's next-generation model Hy4 is about to be released. The platform said it looks forward to the new version's performance on the Agent Arena agent evaluation leaderboard.
Data published by Arena.ai shows that the Hy4 preview version previously released by Tencent Hunyuan has already achieved good results on the Agent Arena leaderboard. In an evaluation of more than 14,500 real-world agent task sessions, the Hy4 preview version ranked second among open models and tenth on the global overall leaderboard, behind only the open model Kimi K3 (Max).
Specifically, the Hy4 preview version's Net Improvement rate reached 5.4%, while Kimi K3 (Max) reached 6.6%, a difference of 1.2 percentage points. This metric measures the net effect improvement a model brings in real agent tasks.
Notably, the Tencent Hunyuan Hy4 preview version shows a certain advantage in cost. Arena.ai data shows that the median cost for the Hy4 preview version to complete a single task is about $0.26, while that for Kimi K3 (Max) is $0.80, making the former about 68% lower in cost.
With its balance between performance and cost, the Hy4 preview version has entered the Pareto Frontier of Agent Arena, meaning it has strong performance and cost competitiveness among currently evaluated models.
In terms of specific capability metrics, Arena.ai disclosed that the Hy4 preview version's performance in dimensions including Confirmed Success, Praise vs. Complaint, and Bash Recovery was 12%, 9%, and 7.5%, respectively. These metrics reflect the model's performance in task execution effectiveness, user feedback, and recovery from tool-call anomalies.
In addition, the evaluation found no Tool Hallucination issue in the Hy4 preview version, that is, the model calling tools that do not actually exist.
By task category, among open models, the Hy4 preview version ranked second in both the Code and Work categories, and third in the Chat category.
Arena.ai believes that the Hy4 preview version achieves competitive agent task performance at a relatively low usage cost, which is positive for the open-source model ecosystem.
At present, Tencent Hunyuan has not announced the specific release time, model parameters, or complete technical details of the official Hy4 version. Whether the new version can further improve agent task performance and whether it will continue the preview version's cost advantage remains to be verified after the official release. (Editor/Yu Lei)
For more first-hand news, welcome to download the Phoenix News app and subscribe to Phoenix Tech. For in-depth reports, please search "Phoenix Tech" on WeChat.