A Chinese Lab Just Launched a Free Rival to Claude Code — At a Tenth of the Price
Z.ai (formerly Zhipu AI) launched ZCode, a free coding agent built on its GLM-5.2 model, priced at roughly a tenth of Anthropic's Claude Code and Max tiers.
If you’ve used Claude Code or OpenAI’s Codex to let an AI write and run code for you, there’s a new, much cheaper option in the ring: Z.ai, the Beijing lab formerly known as Zhipu AI, just launched ZCode — a free desktop app that does the same job, built around its own model, GLM-5.2.
A coding agent isn’t a chatbot you copy-paste into. It’s an assistant that can read your files, run terminal commands, browse related code, and make Git changes on its own while you supervise. ZCode works the same way Claude Code and Codex do: one agent, one workflow, handling the whole loop. It’s free to download for macOS, Windows, and Linux, supports “bring your own key” so you can plug in other models if you want, and — in a twist you won’t find in the US tools — it can be controlled remotely through WeChat or Feishu, or from your phone.
The pricing is the real story. GLM-5.2, the model behind ZCode, is available under Z.ai’s GLM Coding Plan, described as costing roughly a tenth of what Anthropic charges for its comparable Claude Code and Claude Max tiers. GLM-5.2 itself shipped in June under an MIT license — one of the more permissive licenses out there — and it’s not just cheap talk: an independent comparison by Snowflake across 103 coding tasks found GLM-5.2 and Anthropic’s Opus 4.7 nearly tied after three attempts each.
What’s actually going on: Building a frontier-class coding model used to be something only two or three labs could pull off. That’s changing fast. Chinese labs — Z.ai, DeepSeek, Alibaba’s Qwen team, and others — have been closing the capability gap while undercutting Western pricing by an order of magnitude. For a while, “good enough, much cheaper” open models were mainly a research curiosity. ZCode is a sign they’re now packaged as polished, everyday tools competing directly for the same desks Claude Code and Codex sit on.
What this means for you: if you’re new to AI coding tools, the concept hasn’t changed — you still get an assistant that writes and tests code with your oversight, just from a different company. If you already use Claude Code or Codex and cost matters to you, ZCode is worth a trial run, especially for side projects; just know you’re sending your code to Chinese servers unless you self-host via BYOK, which matters if you handle sensitive client work. For most casual users: nothing you need to switch today, but the price pressure this creates will likely show up in what you pay for these tools going forward, regardless of which one you pick.
Sources
AI research agents don't fail at searching — they fail at asking you questions
A new benchmark from Tencent and Tsinghua finds that when questions are ambiguous, AI agents that search harder do worse than ones that simply ask the user — and even the best model stays below 50 percent.
