GLM 5.1🔒Private

• 🔒 Runs in a Trusted Execution Environment with hardware attestation.
• 🧠 Post-training upgrade to GLM-5 for agentic engineering work.
• 📏 200,000-token context window, FP8 precision.
• 🏢 Large Mixture-of-Experts design per NVIDIA's model card.
• ⚡ Z.ai documents up to 8 hours of autonomous task execution.
• 🔧 Reasoning, web search, and tool-calling agent capabilities.
• 📚 Open weights under the permissive MIT license.
• 🌐 From Z.ai (formerly Zhipu AI), released April 2026.

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 5.1 is the privacy-hardened deployment of Z.ai's long-horizon flagship, served inside a Trusted Execution Environment where hardware attestation lets users independently verify enclave identity and configuration, so even the host cannot read prompts. It runs in FP8 precision under the MIT license, with a 200,000-token context window, and is released by Z.ai (formerly Zhipu AI) in April 2026.

It builds directly on Z.ai's GLM-5 foundation as a post-training upgrade rather than a new architecture. NVIDIA's model card describes GLM-5.1 as a large Mixture-of-Experts model and states it has significantly stronger coding capabilities along with improved results across coding, math, and agentic tasks versus its same-family predecessor GLM 5.

Z.ai positions the model for sustained autonomous work. Its documentation describes planning, execution, testing, and fixing loops running up to eight hours, and reports a 3.6× geometric-mean speedup on KernelBench Level 3 optimization, compared with 1.49× for torch.compile in max-autotune mode. These figures are vendor-reported.

Within the broader lineage GLM 5.1 sits between GLM 4.7 and the newer, longer-context GLM 5.2, reflecting Z.ai's steady iteration on its open-weight GLM family.