Last updated: 2026-07-06T13:18:52.272Z

Z.ai Review: Inside GLM-5.2 and ZCode, China's Answer to Cursor and Claude Code

Z.ai, the Beijing-based lab behind the GLM model family, just launched ZCode — a free, agent-first coding environment built on GLM-5.2. Here's what it does, how it's priced, and how it stacks up against Western coding agents.

Z.ai Review: Inside GLM-5.2 and ZCode, China's Answer to Cursor and Claude Code

Chinese AI lab Z.ai has quickly become one of the most talked-about names in the AI coding space in 2026. Formerly known as Zhipu AI, the company rebranded internationally as Z.ai in 2025 and has spent the past year releasing a rapid string of open-weight models under the GLM (General Language Model) family — culminating in its most ambitious product yet: ZCode, an AI coding agent built to compete directly with Cursor, GitHub Copilot, and Claude Code.

You can find Z.ai listed on our directory here: Z.ai on Nexorzo.

Who Is Z.ai?

Z.ai's flagship product is the GLM family of large language models, which the company has open-sourced under the MIT License since July 2025. The company traces its roots to Tsinghua University, having been founded in 2019 before spinning out as an independent company. It is considered one of China's "AI tiger" companies and the third-largest LLM player in China's AI industry, according to IDC.

The company has had a turbulent but high-profile run. In January 2025, the U.S. Commerce Department added the company to its Entity List over national security concerns. Despite that, Z.ai has continued releasing models at a breakneck pace, and its business momentum has only accelerated. As of the IPO process, the company was headquartered in Beijing with roughly 127 employees. Following the launch of its latest model, Zhipu AI's market capitalization crossed HK$1 trillion (around US$128 billion) after a 42% intraday share surge, and JPMorgan raised its 2026–2030 revenue forecasts for the company by 7–16%, projecting over 534% revenue growth for 2026 alone.

GLM-5.2: The Model Behind the Hype

Z.ai's newest flagship model, GLM-5.2, launched in mid-June 2026 and immediately made waves. The launch was compared by some in Silicon Valley to a second "DeepSeek moment" for Chinese open-source AI — echoing the disruption caused by DeepSeek the previous year. Benchmarks published around launch claimed GLM-5.2 outperformed some leading Western frontier models on key coding and reasoning tests, fueling both excitement and skepticism in equal measure.

The model builds on GLM-5.1, which itself made headlines for stamina: GLM-5.1 was reported capable of working autonomously on a single task for up to 8 hours. Z.ai has leaned heavily into this "long-horizon autonomous agent" positioning rather than competing purely on chat quality.

ZCode: Z.ai's Coding Agent

The centerpiece of Z.ai's current push is ZCode, launched in early July 2026 as an "Agentic Development Environment" built around GLM-5.2, shipping as desktop apps for macOS, Windows, and Linux. Z.ai describes ZCode as a "harness" — a control system that lets the underlying GLM-5.2 model execute multi-step coding tasks autonomously, functioning as a true AI agent rather than a simple autocomplete tool.

A few things stand out about ZCode's design:

- Aggressive, undercutting pricing — ZCode ships free to download, with paid GLM Coding Plan tiers layered on top: Lite at $16.20/month, Pro at $64.80/month with 5x the usage quota and MCP tool access, and Max at $144/month with 20x the quota — undercutting comparable US tools that typically run $20 to $200/month at similar tiers.
- Remote, chat-based control — users can trigger and steer coding tasks remotely through WeChat, Feishu, or Telegram, rather than requiring a desktop session. This chat-first workflow is a genuinely different distribution strategy compared to Western IDE-centric tools.
- BYOK support — ZCode supports bring-your-own-key setups for third-party models, and bundles more than 20 programming tools including Git and a terminal, organizing work through a "Goal" construct that spans planning, execution, and verification.
- Fast ecosystem adoption — according to Z.ai's own announcements, major IDEs and developer platforms, including TRAE, Qoder, OpenCode, and WPS Office, integrated GLM-5 within 24 hours of its release, signaling unusually quick third-party buy-in.

To sweeten the launch, Z.ai rolled out promotions increasing data quotas for existing subscribers by 50% and offering 5 million free tokens to new ZCode users.

How Does It Compare to Claude Code and Cursor?

Z.ai is entering a crowded, fast-moving market. Enterprise AI coding agents were estimated at roughly $9.8–11.0 billion in annualized market size as of April 2026, according to Gartner, with frontier model providers increasingly competing directly against application-layer coding tools — exactly the pattern ZCode represents. For context on scale, Cursor has reportedly reached $2 billion in ARR, while Claude Code hit approximately $2.5 billion in annualized revenue by early 2026.

Notably, Z.ai was not among the vendors evaluated in Gartner's 2026 Magic Quadrant for Enterprise AI Coding Agents, which named Anthropic, Cursor, GitHub, and OpenAI as Leaders — a reflection of Z.ai's still-limited enterprise sales presence outside China rather than a judgment on the underlying technology.

The Real Consideration: Data Governance

Beyond price and performance, there's a less obvious factor worth flagging for any team evaluating ZCode seriously: where your code and prompts actually go. Adopting a coding agent built by a Chinese lab means routing source code and prompts through infrastructure subject to different data-governance and export-control assumptions than US-based tools. While BYOK support lets teams keep model inference on their own infrastructure, ZCode's orchestration, task history, and chat-based triggers still run through Z.ai's own systems. For hobby projects, that's a non-issue. For enterprise codebases, it's worth treating as a first-order procurement question rather than an afterthought.

Should You Try Z.ai?

If you're a developer curious about fast-moving, low-cost alternatives to Western coding agents — or you just want to see what a genuinely different distribution model (chat-app-triggered coding tasks) feels like — ZCode is worth a look, especially given its free desktop tier. For production, enterprise-grade work involving sensitive code, it's reasonable to wait and watch how the data-governance conversation develops before committing fully.

You can explore Z.ai and see how it fits into the broader AI landscape on our directory: Z.ai on Nexorzo.

---

Looking for more AI tools like this? Browse our full collection at Nexorzo's AI Tools Directory.