source: simon willison: glm-5.2 is probably the most powerful text-only open weights llm
level: technical
chinese ai lab z.ai released glm-5.2 under an mit license on june 16, 2026. the model has 753 billion total parameters with 40 active via mixture of experts, making it a 1.51tb text-only model. it expands the context window to 1 million tokens from the previous 200,000. glm-5.2 is available through openrouter from nine providers, with typical pricing at $1.40 per million input tokens and $4.40 per million output tokens.
independent benchmarker artificial analysis ranks glm-5.2 as the leading open weights model on its intelligence index v4.1, scoring 51. this places it ahead of minimax-m3 at 44, deepseek v4 pro at 44, and kimi k2.6 at 43. however, the model uses more output tokens per task—43,000 on average—compared to its predecessor glm-5.1 at 26,000 and competitors like minimax-m3 at 24,000. on the code arena webdev leaderboard for front-end tasks, it ranks second only to claude fable 5, despite lacking image input.
in qualitative tests, glm-5.2 produced a well-animated svg of a pelican riding a bicycle, with smooth, intact animations. but it fell short on a prompt for a north virginia opossum on an e-scooter, delivering a static, less detailed image compared to the animated version from glm-5.1. this mixed performance highlights both the model's strengths in certain creative tasks and its inconsistency across different prompts.
why it matters: glm-5.2 sets a new bar for open weights text models, offering strong reasoning and coding at lower cost than proprietary alternatives, though its high token usage may affect efficiency.
source: simon willison: glm-5.2 is probably the most powerful text-only open weights llm