Zhipu releases GLM-4.7 for agentic coding

Background

Zhipu released GLM-4.7 in December 2025 alongside the GLM-4.6V vision model.^[2][3]

What happened

Zhipu's GLM-4.7 is aimed at multilingual agentic coding and terminal tasks, which means the model writes and changes code over several steps and works in a command-line environment. The weights were posted to Hugging Face on 22 December 2025.^[2][3]

Key details

On SWE-bench Verified, which tests whether a model can resolve real issues in open-source software, GLM-4.7 scored 73.8%, 5.8 points above GLM-4.6. On SWE-bench Multilingual it scored 66.7%, 12.9 points higher.^[1]

On Terminal Bench 2.0 it scored 41.0%, up 16.5 points, and on HLE with tools it scored 42.8%, up 12.4 points.^[1] The model adds Interleaved Thinking, which thinks before each response and each tool call, and Preserved Thinking, which keeps reasoning across multi-turn agent sessions. Turn-level thinking control lets users manage thinking turn by turn.^[1]

Impact

GLM-4.7 is released under the MIT License, so developers can download it and build agent tools on it under the terms of that licence.^[1]

Reading the benchmark figures

A benchmark score is only meaningful when two models are run on the same test under the same conditions. Vendors usually publish comparison tables that set their model against selected rivals on selected tests, so a number should be read together with the name of the test and the models it is compared with. A small gap between two scores does not by itself show a real difference in ability, and one test rarely captures everything a model can do.

Agentic terms

An agent is a model that works through a task in steps, deciding what to do next and calling external tools such as search or code execution along the way. Thinking modes control whether and how a model reasons before it answers. A context window is the amount of text a model can use at one time, and a longer window lets a model keep more of a long session or document in view.