Tencent has expanded inference capacity for its newly released Hy4 model tenfold following a surge in usage on WorkBuddy. Announced on November 5, the expansion comes a day after the model’s launch, alongside benchmark results highlighting its general intelligence, agent capabilities and cost efficiency.
Hy4 first became available through WorkBuddy, where strong initial demand led many users to reach their usage limits on the first day. Tencent has since increased the capacity of its Hy4 inference cluster and raised free usage allowances, while continuing to adjust resource allocation to accommodate demand.
The expansion coincides with new results from independent AI evaluations. In the November 5 update to the Artificial Analysis Intelligence Index, Hy4 ranked second among Chinese models, behind Kimi K3.1 Pro and ahead of Qwen 3.8 Max and DeepSeek V4.1 Flash. It placed among the leading entries in a field of 724 models, matching the score of Grok 4.7 (xhigh).
The index assesses performance across a range of demanding tasks, providing a broader measure of model capability than a single benchmark. Hy4’s result places it alongside leading models on overall performance, with a substantially lower cost per task than several similarly scoring alternatives.
According to the figures cited in Tencent’s announcement, Hy4’s cost per task in the Artificial Analysis evaluation was approximately US$0.78—around one-tenth that of Grok 4.7 (xhigh) and one-thirteenth that of Qwen 3.8 Max. These figures reflect costs within the evaluation rather than a flat price for every user task.
On Agent Arena, which evaluates models through real-world agent interactions, Tencent reported that Hy4 ranked 13th overall and first among open-source models. The model also appeared on the leaderboard’s Pareto frontier, identifying it as one of the leading options for its combination of performance and cost.
Hy4 maintained a positive net improvement score while operating at a substantially lower cost than higher-priced models such as Claude Opus 5.5. The result adds a practical agent evaluation to the broader capability assessment provided by Artificial Analysis.
Released on November 4, Hy4 is Tencent Hunyuan’s latest-generation large language model. Compared with Hy4 preview, it was developed with substantially expanded data and training scale across both pre-training and post-training. The release introduces visual understanding for the first time, alongside improvements in instruction following, long-context comprehension and long-horizon agent tasks.