Z.ai opens the 744B parameter GLM-5.1 to all coding subscription users, starting at approximately $10/month.

Headline

Z.ai is opening GLM-5.1 to all coding subscription users

Overview

Z.ai (with a Tsinghua-affiliated background, founded in 2019, listed on the Hong Kong Stock Exchange in January 2026) announced that its 744B-parameter GLM-5.1 is now available to all coding subscription tiers. This is a model based on MoE architecture.

Billing is quarterly: Lite $27, Pro $81, Max $216, which amounts to roughly $10 per month at the minimum. Z.ai claims that compared to Anthropic’s Claude Pro, at the same price tier, users get 3 to 5 times more quota. It can directly connect to Claude Code and Cursor.

Its SWE-bench Verified score is 77.8%, higher than Gemini 3 Pro, and comparable to Claude Opus 4.5. The model is open-sourced under the MIT license.

Analysis

• The price war intention is very clear:

  • Low prices combined with higher quotas are very attractive for long-term coding scenarios. The uncertainty of billing based on tokens has always been a pain point; fixed subscriptions can solve this issue.
  • Opening to “all subscription tiers” indicates they want to rapidly expand their developer base.

• The compute supply chain is a selling point:

  • Z.ai uses Huawei Ascend chips for inference. Against the background of US export restrictions, having an independent supply chain is attractive to enterprise users worried about compute supply disruptions.

• Early feedback on user experience:

  • Some developers say hallucination rates in coding tasks are lower than GPT-4.5, but the sample size is still insufficient to draw definitive conclusions.
  • Fixed subscriptions are more friendly for team collaboration and scenarios involving continuous agent runs—they eliminate the need to constantly monitor token consumption.

• How competitors might respond:

  • In the short term, Anthropic and OpenAI are unlikely to follow with price cuts; they are more likely to emphasize differences in quality and stability. However, GLM-5.1’s scores, along with its ability to connect to Claude Code and Cursor, make it attractive to cost-conscious developers.

Impact assessment

• Importance: High
• Category: Model release / Developer tools / Market landscape

Conclusion: If you are building heavy coding agents or often run long sessions, it’s worth trying now—the cost advantage is very direct. For short-term trading where relevant underlying assets cannot be obtained, the marginal gains are limited. In the long run, we still need to observe whether quality and ecosystem stickiness can be maintained and continue to compete with top-tier closed-source models.