Overview
GLM-5.2 is Z.ai's new flagship coding model for the GLM Coding Plan, announced on June 13, 2026 and rolled out to Lite, Pro, Max, and Team users. It is positioned as the upgrade path for complex coding-agent work: large repositories, long-running sessions, deeper reasoning, and workflows that need a usable 1M-token context window rather than the 200K-class context used by earlier GLM-5.x models.
The practical change is not just a model name. Z.ai documents glm-5.2[1m] specifically for Claude Code 1M-context setup; OpenClaw and Cline-style custom providers use glm-5.2 with the documented 1M context-window setting where supported. The [1m] suffix enables the 1M-context variant when paired with the CLAUDE_CODE_AUTO_COMPACT_WINDOW=1000000 setting in Claude Code. For teams already using Z.ai's Coding Plan inside supported coding tools, GLM-5.2 is the version to test on the hardest agentic engineering tasks.
This release should still be treated carefully. Z.ai's current documentation emphasizes availability, setup, context length, effort mapping, and quota rules; it does not publish a full GLM-5.2 benchmark table in the docs reviewed here. API, chatbot, and open-source availability were announced as planned for the week after the June 13, 2026 release, so this page avoids treating those channels as already live until Z.ai publishes the corresponding docs or repository.
What's New
GLM-5.2 for All Coding Plan Tiers
The GLM Coding Plan now lists GLM-5.2 as supported across Lite, Pro, Max, and Team plans. That is a wider rollout than an invite-only or Max-only release: existing Coding Plan users can switch their supported coding agent to GLM-5.2 without waiting for a separate enterprise enablement path.
Z.ai describes GLM-5.2 as an advanced model comparable to Claude Opus-class usage inside the Coding Plan. That makes it a better fit for complex tasks than routine autocomplete or small bug fixes, where the documentation still recommends GLM-4.7 to preserve quota.
1M-Context Mode
GLM-5.2 adds a documented 1M-context path through the glm-5.2[1m] model identifier. In Claude Code, Z.ai's setup example pairs that model name with:
{
"env": {
"CLAUDE_CODE_AUTO_COMPACT_WINDOW": "1000000",
"ANTHROPIC_DEFAULT_HAIKU_MODEL": "glm-4.5-air",
"ANTHROPIC_DEFAULT_SONNET_MODEL": "glm-5.2[1m]",
"ANTHROPIC_DEFAULT_OPUS_MODEL": "glm-5.2[1m]"
}
}
For OpenClaw, the documented model entry sets contextWindow to 1000000 and maxTokens to 131072. For Cline-style custom providers, Z.ai recommends using the OpenAI-compatible API provider, setting the base URL to https://api.z.ai/api/coding/paas/v4, entering the Z.ai API key, choosing a custom model such as glm-5.2, disabling image support, and setting context window size to 1000000.
Effort Mapping for Deeper Coding Reasoning
GLM-5.2 exposes effort behavior through Claude Code's /effort command. Z.ai maps Claude Code's low, medium, and high settings to GLM-5.2's high effort, while xhigh, max, and ultracode map to GLM-5.2's max effort. The official guidance is to use max effort for coding tasks when deeper reasoning and more stable complex-task performance matter.
This matters for upgrade testing because a team can mistakenly evaluate GLM-5.2 at a shallower effort level and miss the intended behavior. If your goal is long-horizon planning, multi-file refactoring, or agentic debugging, test with max effort and track quota consumption separately.
OpenClaw and Multi-Tool Configuration
OpenClaw users who cannot switch models through a provider picker can manually add glm-5.2 to ~/.openclaw/openclaw.json, update the primary model from zai/glm-5 to zai/glm-5.2, then restart the gateway. Z.ai also documents custom-provider setup for tools that support OpenAI-compatible endpoints, while warning that tools without custom model settings must wait for official support.
The result is a practical compatibility update for AI coding agents rather than a standalone chat-only release. GLM-5.2 is meant to sit inside developer tools where model routing, context windows, max output, and effort settings directly affect real work.
Compatibility Notes
GLM-5.2 access is currently documented through the GLM Coding Plan and officially supported coding tools. It is not safe to assume every SDK, general chatbot surface, or third-party agent runner can use the model immediately.
Supported or documented paths include Claude Code, OpenClaw, Cline-style OpenAI-compatible custom provider setup, and other coding tools that allow custom model configuration. Z.ai states that unsupported or unauthorized use of Coding Plan benefits may lead to restrictions, so teams should keep GLM-5.2 usage inside approved Coding Plan scenarios until API and chatbot services are formally documented.
The 1M-context path also has a versioning footnote: if Claude Code reports that the [1m] model does not exist, Z.ai recommends upgrading Claude Code and trying again. In practice, rollout issues may be tool-version or configuration problems rather than model unavailability.
Availability & Access
GLM-5.2 is available to GLM Coding Plan users on Lite, Pro, Max, and Team plans. Z.ai's public materials stated that API and chatbot services were planned for the week after the June 13, 2026 release, and that the model was planned for an MIT-licensed open-source release. Until those channels are live in official docs or repositories, the dependable access path is the Coding Plan inside supported coding tools.
For users comparing Claude Code, OpenClaw, Cline, and similar agentic coding workflows, the key question is not only "Can I call the model?" but "Can my tool expose the right model string, context window, and effort settings?" GLM-5.2's most important benefits require correct configuration.
Pricing & Plans
GLM-5.2 is included in the GLM Coding Plan rather than priced as a standalone public API model in the current docs reviewed. Z.ai's Coding Plan overview states that plans start at 18 USD per month and provides estimated usage limits by tier.
| Plan | GLM-5.2 access | 5-hour estimated limit | Weekly estimated limit |
|---|---|---|---|
| Lite | Included | Up to approx. 80 prompts | Up to approx. 400 prompts |
| Pro | Included | Up to approx. 400 prompts | Up to approx. 2,000 prompts |
| Max | Included | Up to approx. 1,600 prompts | Up to approx. 8,000 prompts |
| Team | Included | Team-specific limits and policy | Team-specific limits and policy |
GLM-5.2 and GLM-5-Turbo consume quota faster than routine models. Z.ai says these advanced models are deducted at 3x during peak hours and 2x during off-peak hours, with a limited-time off-peak 1x benefit through the end of September. Peak hours are 14:00-18:00 UTC+8.
For cost control, the official guidance is simple: use GLM-5.2 for complex tasks and continue using GLM-4.7 for routine development work.
Best For
- Developers working on large repositories where 200K context is no longer enough
- Teams running long Claude Code or OpenClaw sessions that need deeper planning and more stable multi-step execution
- Coding Plan users deciding when to spend premium quota on GLM-5.2 instead of GLM-4.7
- Agent builders comparing MCP, tool use, and model-routing behavior across coding environments
- Early adopters who can validate configuration, quota behavior, and output quality before broader API or chatbot rollout
FAQ
What is GLM-5.2?
GLM-5.2 is Z.ai's new flagship coding model for the GLM Coding Plan. It is documented for supported coding tools and adds a 1M-context configuration path for complex coding-agent workflows.
Is GLM-5.2 available to Lite users?
Yes. Z.ai's documentation says all GLM Coding Plan tiers support GLM-5.2, including Lite, Pro, Max, and Team plans.
How do I enable the 1M context window?
In Claude Code, use the glm-5.2[1m] model identifier and set CLAUDE_CODE_AUTO_COMPACT_WINDOW to 1000000. In OpenClaw or Cline-style tools, configure the model/context window according to Z.ai's documentation.
Does GLM-5.2 have public benchmark scores?
The official docs reviewed here do not publish a full GLM-5.2 benchmark table. They position it as a new flagship model with strong coding, 1M-context support, and long-horizon task strengths, but this page avoids inventing specific benchmark numbers.
Is GLM-5.2 available through API or chatbot?
Z.ai's public announcement said API and chatbot services were planned for the week after the June 13, 2026 release. Until official docs confirm those channels, the reliable access path is the GLM Coding Plan inside supported coding tools.
Is GLM-5.2 open source?
Z.ai's public announcement says GLM-5.2 will be open-sourced under the MIT License the following week. This page treats that as planned availability until the repository or official model page is published.
Should I use GLM-5.2 for every coding task?
No. Z.ai recommends GLM-5.2 for complex tasks and GLM-4.7 for routine tasks because GLM-5.2 consumes quota at a higher multiplier.
