Tencent's Hy4 calls have surged and its capacity has been expanded tenfold, ranking first in the open source model of Agent capabilities

On November 5, Tencent announced that due to the surge in Hy4 calls, WorkBuddy has been expanded tenfold to meet the experience and calling needs of more users. At the same time, with the continuous improvement of model capabilities, Hy4 is highly recognized by the international authoritative model testing list. It ranks second among Chinese models on the Artificial Analysis Intelligence Index list and ranks first among open source models on the Agent Arena list.

Since Hunyuan Hy4 was first connected to WorkBuddy, it has received positive experience and enthusiastic feedback from users and developers with its significant improvement in Agent capabilities. On the first day of launch, a large number of users touched the quota. At present, Tencent has expanded the capacity of the Hy4 inference cluster tenfold, increased the free user experience limit, and continued to dynamically allocate resources.

Echoing the enthusiastic response from users, Hy4 simultaneously delivered outstanding results in a number of authoritative overseas reviews. On the 5th, the world's authoritative large model evaluation platform Artificial Analysis released the latest intelligence index list. Tencent Hy4 ranked among the top 724 models, ranking second among Chinese models, second only to Kimi K3.1 Pro, and ahead of Qwen3.8 Max, DeepSeek V4.1 Flash and other models.

Artificial Analysis is regarded by the industry as one of the current important references for measuring the comprehensive intelligence level of large models. This index is not a single running score, but a highly difficult evaluation that comprehensively covers core capabilities such as complex reasoning, code generation and debugging, tool invocation, multi-modal understanding, long-range planning, and scientific problem solving. It requires high generalization capabilities and engineering practicality of the model. Hy4 is tied with Grok4.7 (xhigh) in this comprehensive evaluation system, showing Tencent's solid accumulation of basic model capabilities.

Artificial Analysis also mentioned that in terms of cost efficiency, the single task cost of Hy4 is about US$0.78, which is significantly lower than Grok4.7 (xhigh) and Qwen3.8 Max within the same score range. The price is 10 times and 13 times cheaper than these two models respectively, reflecting a strong capability-cost balance.

In addition, on the latest list of Agent Arena, which focuses more on the performance of real agent tasks, Tencent Hy4 ranked 13th on the overall list, ranking first among the global open source models. With its high cost performance, Hy4 is also located on the Pareto Frontier of this list. Compared with top models such as the more expensive Claude Opus5.5, it significantly reduces costs while still maintaining positive performance.

On November 4, Tencent Hunyuan's latest generation large language model Hy4 was released, which once again fully improved the model's capabilities. Compared with Hy4 preview, Hy4 significantly expanded the data and training scale in pre-training and post-training, introduced visual understanding capabilities for the first time, and greatly improved the comprehensive effects of instruction following, long text understanding, and long-range agents.