GLM 4.7

• 🔒 Runs in a Trusted Execution Environment with hardware attestation
• 🧠 Adds Preserved and Turn-level Thinking for stable reasoning
• 🔧 Flagship tuned for agentic coding and terminal tasks
• 👁️ Improved visual-code understanding and front-end UI aesthetics
• 📏 128,000-token context window in this Venice configuration
• 🌐 Web search capability included alongside reasoning
• 📚 Released under the permissive MIT license
• 🆕 Builds on GLM-4.6's interleaved thinking foundation

Z.ai

Z.ai, formally Knowledge Atlas Technology Joint Stock Co., Ltd., is a Chinese technology company specializing in artificial intelligence. Previously known internationally as Zhipu AI, the company rebranded to Z.ai in 2025. Its core focus is the GLM family of…

GLM 4.7 is Z.ai's flagship text model, here deployed by Venice inside a Trusted Execution Environment so that confidential inference can be independently verified through hardware attestation evidence. The underlying GLM-4.7 model was released by Z.ai (formerly Zhipu AI) under the MIT license, and is positioned for task-oriented development, multi-language coding, and complex multi-step agent workflows.

Compared with its same-family predecessor GLM 4.6, GLM 4.7 extends the interleaved thinking introduced in the GLM-4.5 generation by adding Preserved Thinking and Turn-level Thinking, which keep reasoning state consistent across turns and let the model "think before acting" within coding frameworks like Claude Code, Cline, and Roo Code. Z.ai also highlights stronger visual-code and UI understanding, yielding more consistent layouts and styling for front-end generation.

Within Venice's lineup it sits alongside the lighter GLM 4.7 Flash, and is succeeded by the newer GLM 5.1, which moves to the GLM 5 generation. This GLM 4.7 build pairs reasoning, code optimization, and web search.

For deployment, the open release supports inference frameworks including vLLM and SGLang, and the model card documents its benchmark methodology, including adjustments for tau-squared-Bench user-interaction evaluation. The TEE wrapper here adds an end-to-end-encrypted, attestable execution path on top of those base capabilities.