Tencent says peak-hour limits may persist because its high-end computing is finite, and the WorkBuddy free quota for Hy4 and free Hy3 access both end November 15.
Tencent said on November 5 that it had expanded the Hy4 inference cluster tenfold after many WorkBuddy users reached their quotas on launch day. Inference is the computing that runs a trained model each time someone sends it a request. IT Home carried the announcement the same day, Sina republished a Mydrivers report that credits it to Tencent public relations director Zhang Jun, and TechNode summarized it a day later.
Tencent also raised the free usage allowance for Hy4 and said it will keep adjusting resources as demand shifts. It acknowledged that limits may still appear at some hours, because its high-end computing capacity is finite and use is concentrated at peak times. Users who hit a limit are advised to switch to Hy3, the previous Tencent model, whose free access on WorkBuddy now runs until 23:59 on November 15, or to save non-urgent work for the evening.
Tencent gave no absolute figures: not the size of the cluster before the expansion, the number of affected users or the length of the queues, so the tenfold multiple has no fixed scale. It is the second time in just over two months that a Hy4 release has outrun the computing Tencent set aside for it, which matters to anyone tracking how much capital the company must spend to support its agent products.
Hy4’s launch put WorkBuddy at the center of Tencent’s pitch
Tencent released the official Hy4 on November 4. According to its launch page, the model adds image understanding and a computer-use function for the first time, and it is available on Tencent Cloud’s TokenHub platform, WorkBuddy and CodeBuddy, the Yuanbao assistant, the WeCom assistant and the ima app. IT Home reported that WorkBuddy was the first product to integrate it.
The launch page leans on WorkBuddy for evidence. Tencent reports that Hy4 scored an average of 83.18 and passed 96.15% of tasks in WorkBuddy’s own evaluation, ranking first among Chinese models in nine of 12 scenarios. Those are Tencent’s figures and carry no outside verification. TechNode, citing Tencent’s materials, adds that Hy4 ranks second among Chinese models on the Artificial Analysis intelligence index, behind Kimi K3.1 Pro, and 13th overall on Agent Arena.
WorkBuddy already has a large corporate base to serve. On October 29, Tencent said the enterprise edition covers half of China’s 500 largest companies, as EastFrontier reported at the time. EastFrontier’s coverage of the Hy4 release details what the model can do.
Free quotas and scarce chips turn launch demand into a capacity bill
Tencent is steering demand with price. The launch page offers a set free quota on WorkBuddy and CodeBuddy from November 4 to November 15, and it bills Hy4 on Tencent Cloud’s TokenHub platform at 50% of list price through December 31. The list price is RMB6 per million input tokens and RMB18 per million output tokens, unchanged from the Hy4 preview. A cache hit, where the model reuses text it has already processed, costs RMB0.3 per million tokens.
A free window invites heavy trial use, and this is not the first time it has produced a queue. Caiwen News, as republished by Tonghuashun on August 31, reported that the Hy4 preview was first integrated into WorkBuddy on August 28, that queues formed in the task line on the first day, and that Hunyuan apologized and urgently expanded the preview cluster. Tencent’s wording then matched its wording now: it cited limited high-end computing capacity and concentrated peak demand as reasons that queues could return.
Tencent has not said which hardware limits it. The November notice offers a stopgap instead of a date for relief: send overflow traffic to Hy3, the older model, and shift non-urgent jobs to the evening.
The November 15 deadline will show whether demand holds once the free quotas end
The free quota and the Hy3 extension both end on November 15, so Tencent will soon see how many WorkBuddy users stay on Hy4 once the free quota ends. Tencent has not said whether it will keep the raised quotas after that date.
Two other promises could change the load. Tencent says it will open-source the Hy4 model weights soon and will publish a system card describing its security testing, and a lower-cost Hy4 Flash is also listed as coming soon. A cheaper model would give Tencent another place to route overflow demand.
The next signal is Tencent’s own: a statement of quota rules after November 15, a release date for Hy4 Flash, or another queue notice on WorkBuddy.