Zhipu AI ZCode launch: GLM-5.2 coding agent vs Claude Code

Zhipu AI has launched ZCode, a browser-based coding agent built around GLM-5.2 that lets users write, debug, test and review code in a single workflow. ZCode exposes a dedicated agent that handles tasks, file access, terminal output, browser context and Git changes, and it supports a 1M-token context window for multi-step programming tasks.

What is ZCode and how does it work?

ZCode is a coding agent that runs on GLM-5.2 and orchestrates coding tasks across files, terminals, a browser context and Git, all within a single workflow. The agent lets users give natural language instructions to write, debug, test and review code; Z.ai demonstrates it building a browser-based Gomoku game with integrated Git tools and task tracking.

The agent can be controlled remotely via Feishu, WeChat or a smartphone, and Z.ai says the 1M-token context window allows even multi-step programming tasks without losing context. ZCode packages the language model, file and terminal access, browser state and Git operations into one flow rather than juggling separate tools.

How does ZCode compare on performance and price?

GLM-5.2, the model powering ZCode, shipped in June 2026 under an MIT license, and Z.ai positions it as competitive with pricier Western models while costing less. A hands-on comparison by Snowflake across 103 tasks shows GLM-5.2 and Opus 4.7 nearly tied after three attempts, a result the company highlights when contrasting GLM-5.2 with models like Claude Opus.

The announcement stresses cost advantage rather than publishing list prices in this release; it notes that GLM-5.2 has quickly built a following among developers who see it as competitive with pricier Western models like Claude Opus at a fraction of the cost. Z.ai also offers a trial and subscriber quotas: new customers get a free five-day trial with up to 5 million tokens per day, and subscribers receive about 1.5 times more quota through July 2026.

How will developers use ZCode day to day?

ZCode aims to centralize typical developer workflows by handling code editing, testing, debugging and version control through natural language prompts. That means a single agent can propose Git changes, run tests and show terminal output in context, reducing the back-and-forth between separate windows and human instructions.

Z.ai’s example—a browser-based Gomoku game—illustrates how the tool ties a development task to Git operations and task tracking inside the same session. Remote control via popular messaging platforms or a smartphone suggests Z.ai expects developers to run or inspect agent sessions from multiple devices.

Why it matters

ZCode packages a large-context language model and tool access into one agent, which shortens the path from intent to executed code. The 1M-token context window lets the agent maintain state across long, multi-step programming jobs, and the trial terms make it easy for developers to test that claim: five days with up to 5 million tokens per day. If GLM-5.2 holds parity with Opus 4.7 on real tasks, ZCode could lower the cost and friction of model-assisted development for teams that need sustained context and integrated Git workflows.

What to watch

Watch adoption signals: how quickly developers move from trial to paid subscriptions under the temporary 1.5x quota boost that runs through July 2026, and whether independent evaluations beyond Snowflake reproduce the reported tie between GLM-5.2 and Opus 4.7 across diverse coding tasks.

Source-attributed facts

• GLM-5.2 shipped in June 2026 under an MIT license.
• ZCode uses a 1M-token context window.
• New customers receive a free five-day trial with up to 5 million tokens per day.
• Subscribers get about 1.5 times more quota through July 2026.
• A hands-on comparison by Snowflake across 103 tasks shows GLM-5.2 and Opus 4.7 nearly tied after three attempts.