SINGAPORE – Media OutReach Newswire – 26 December 2025 – Z.ai has launched GLM-4.7, the latest iteration of its open-source large language model, just ahead of Christmas. This release reflects the company’s enhanced focus on equipping its models for real-world software development and production environments.
The new model is engineered to support practical engineering workflows, emphasizing long-running task execution, stable tool calling, and multi-step reasoning. These capabilities are increasingly critical as developers integrate large language models into complex, agent-based systems.
Compared to its predecessor, GLM-4.6, the new model shows significant improvements in code generation, complex reasoning, and agent execution. According to Z.ai, GLM-4.7 provides more consistent and controllable performance during extended tasks, alongside clearer and more concise language output, addressing a common limitation seen in various open-source models.
To assess GLM-4.7’s capabilities in realistic settings, Z.ai evaluated it against 100 practical programming tasks in production-like environments, such as Claude Code, covering front-end, back-end, and command-execution scenarios. The company stated that GLM-4.7 realized higher task completion rates and improved stability than its predecessor, GLM-4.6, and has since been designated as the default model for Z.ai’s GLM Coding Plan.
Benchmark results further position GLM-4.7 among the top open-source models available today. It achieved a score of 67.5 on BrowseComp and 87.4 on τ²-Bench, the latter representing a new peak for open-source systems. In coding-focused assessments, including SWE-bench Verified and LiveCodeBench v6, GLM-4.7’s overall performance closely approaches that of Claude Sonnet 4.5. Additionally, in Code Arena’s extensive blind evaluation, which aggregates votes from over one million comparisons, GLM-4.7 ranked first among open-source models.
The model is accessible through the BigModel.cn API and has been integrated into Z.ai’s full-stack development platform. As open-source models continue to gain prominence in the global technology landscape, Z.ai’s advancements offer a clear indication of how these systems are likely to evolve and what new functionalities they may enable in the future.
For further information, users can explore the model via the GLM Coding Plan at https://z.ai/subscribe, or try it now at https://chat.z.ai/. The weights for GLM-4.7 can be found at https://huggingface.co/zai-org/GLM-4.7, and more technical details are available on Z.ai’s blog at https://z.ai/blog/glm-4.7.
The issuer is solely responsible for the content of this announcement.
See also
Top 7 AI Note-Taking Apps of 2026: Features, Pricing, and User Insights
Top AI Tools for Model Interpretability: ELI5, LIME, SHAP, and More Explained
Dark Web AI Tools Empower Criminals: New Threats to Security and Society
Google Completes December 2025 Core Update, Triggering Major Ranking Volatility
AI Adoption Soars in India, Yet Productivity Gains Remain Invisible in Official Stats
