Term
GLM-4.7
GLM-4.7 (December 2025) is Zhipu AIs open-weight coding flagship — an MoE model with roughly 400B parameters, a 200K input context and the "Preserved Thinking" feature.
GLM-4.7 — explained in more detail
Zhipu AI (Z.ai) released GLM-4.7 on December 22, 2025 as an open-weight model, designed for deep mathematical reasoning, multi-file software engineering and stable agentic orchestration. It uses a Mixture-of-Experts (MoE) architecture with roughly 400B parameters focused on logic and code, supports an input context of 200K tokens and an output length of 128K tokens.
Example / Practical use
New in GLM-4.7 is the Preserved Thinking feature: reasoning chains are maintained across multiple turns instead of being reset at each step — useful for longer agentic workflows. The model is available through the BigModel.cn API, the z.ai development environment, as a local download on Hugging Face, and via OpenRouter and partner platforms such as TRAE, Cerebras, YouWare, Vercel and CodeBuddy. GLM-4.7 was positioned as a low-cost alternative to expensive Western coding tools.
Distinction
GLM is the model range of the Chinese company Zhipu AI and, with version 4.7, marks a shift from pure conversational competency toward agentic reliability. As an open-weight model it differs from proprietary coding models of Western vendors through free access; compared with later, larger models such as the GLM 5 range, 4.7 is the leaner predecessor generation.
Discover more
GPT-6 Astra Found Questions My First Security Review Missed
I used GPT-6 Astra and Fable 5.1 as independent reviewers for RLS, API and tenant-isolation checks. The useful part was the structured cross-review.
GlossaryGLM-5.3
GLM-5.3 is Zhipu AIs open-weights model from August 2026 — a MoE model focused on coding and cybersecurity that tops the open-model field on agentic benchmarks.
EncyclopediaGLM (Zhipu AI / Z.ai) — the open model line at a glance
GLM by Zhipu AI as an open line: positioning, use profile and how to choose. Flagship GLM-5.3 with strong coding performance.