

Z Ai · 2× · tolest 19. Aug. 2026
Z.ai (formerly Zhipu AI, a Chinese company and one of China's "six AI tigers") is the provider of the GLM model series of frontier-capable language models. The current flagship model GLM-5.2 is a Mixture-of-Experts model with approximately 744-753 billion total parameters (around 40 billion active parameters per Token), developed for Coding, Reasoning, and long-context agent tasks. The weights are released under the open MIT license, enabling self-hosting and commercial use. The model is characterized by a usable 1-million-Token context window and is provided via the Z.ai API, the GLM Coding Plan, and third-party providers (Hugging Face, OpenRouter, among others).
Features
| Key Benchmark (%) | GPQA Diamond: 89.5%; τ²-Bench: 99.1%; trails Claude Opus 4.8 by just 1% on FrontierSWE |
| Context Window (Tokens) | 1,000,000 tokens (1M variant), up to 131,072 output tokens max |
| License | MIT License (open weights) |
| Multimodality | Primarily text-based (GLM-5.2); separate vision models such as GLM-4.5V/GLM-4.6V available for image/video |
| Platform | Z.ai API, GLM Coding Plan, Z.ai chatbot, Hugging Face, OpenRouter, third-party providers (Groq, Together, Fireworks, etc.) |
| Price per 1M Tokens | $1.40 input / $4.40 output (standard); $0.26 cached input |
| Release Date | June 13, 2026 (Coding Plan); open weights/API from June 16, 2026 |