GLM-5.2 tops open-weights leaderboard, matches GPT-5.5 on agentic benchmark

GLM-5.2 is the new leading open weights model on Artificial Analysis

Z.ai's GLM-5.2 scores 51 on the Artificial Analysis Intelligence Index v4.1, ahead of MiniMax-M3 (44) and DeepSeek V4 Pro (44), making it the top open-weights model. It keeps the same 744B-total / 40B-active parameter count as GLM-5.1 but posts big gains in scientific reasoning and agentic tasks—HLE jumps 12 points to 40%, CritPt up 16 points to 21%. On GDPval-AA v2, a real-world agent benchmark, it hits 1524, effectively level with GPT-5.5 (xhigh reasoning). The trade-off: it averages 43k output tokens per task, up from 26k on GLM-5.1. API pricing stays at $1.4/$4.4/$0.26 per 1M input/output/cache-hit tokens, context window expands from 200K to 1M, and it ships under an MIT license.

Why it matters: GLM-5.2 hits 51 on Artificial Analysis's Intelligence Index, passing MiniMax-M3 and DeepSeek V4 Pro to become the top open-weights model. Same architecture, +11 points, same pricing. Score capped at 82 because it's a single-benchmark claim from one evaluator—no cross-source co...