Zhipu's ZCode hits 1M users, resets all usage limits
Z.ai says its agentic coding environment passed one million users and wiped usage counters for every GLM Coding Plan subscriber as a thank-you.
A scale marker for Chinese coding agents
Z.ai, the international arm of Beijing-based Zhipu AI, said on Tuesday morning Beijing time that ZCode — its own agentic coding environment built around the GLM model family — has passed one million users. To mark the milestone, the company reset usage limits for every subscriber on its GLM Coding Plan, effectively handing the entire paying base a fresh quota mid-cycle. Z.ai said it is also rolling out an update aimed at converting the models' long-horizon capabilities into finished work rather than long-running attempts.
ZCode is Z.ai's answer to Claude Code, Cursor and GitHub Copilot: a terminal-and-editor harness that drives the company's GLM models through multi-step engineering tasks, with bring-your-own-key support for third-party models and a usage-quota bonus for GLM Coding Plan subscribers. The plan itself has been Zhipu's most successful international product, priced far below Western equivalents and marketed directly at developers looking to migrate off subscription tiers from US labs.
Why the reset is the interesting part
The user number is notable, but the limit reset carries more information. In January, Zhipu capped new sign-ups to the GLM Coding Plan, citing computing constraints as demand outran the serving capacity it could allocate — an unusual public admission from a lab that had been aggressively courting overseas developers. Handing out additional quota seven months later implies the inference bottleneck has eased, whether through better serving efficiency, more accelerators, or both.
That matters beyond one company. Coding agents have become the primary commercial wedge for Chinese labs outside China: they monetise in hard currency, they do not require a consumer app, and they are relatively insensitive to brand. A million users on a single Chinese-built harness — plus the ability to give away rather than ration capacity — suggests the constraint on Chinese open-weight labs competing internationally is shifting from compute supply toward product and trust. It also sharpens the pricing squeeze on Western coding-agent subscriptions, which continue to meter usage tightly while a MIT-licensed alternative periodically zeroes the meter.