GLM-5.2: Z.ai Ships 1M-Token Coding Model With Zero Benchmarks

Short summary

Z.ai released GLM-5.2 with a 1-million-token context window (vs 200K in GLM-5.1) and 131K max output tokens, matching Claude Opus and Gemini on context length. The model adds dual thinking-effort modes (High/Max) for reasoning depth but shipped without published benchmarks—Zhipu is letting developers test on their own codebases before releasing community results. Setup requires three config lines; an OpenAI-compatible API endpoint is live for coding plan subscribers, with public API launching June 16–20.

• •GLM-5.2 now supports 1M-token context (10x increase from GLM-5.1's 200K)
• •Two reasoning modes added: High (default) and Max (extended reasoning for complex tasks)
• •Benchmarks withheld at launch; community will validate performance on real codebases before public API release

Generated with AI, which can make mistakes.