GLM-5.2: Hits #3 on GDPval-AA Agentic Benchmark

GLM-5.2 from Zai_org posted 1524 Elo on the GDPval-AA real-world agentic benchmark, landing at #3 overall behind Claude Fable 5 and Claude Opus 4.8 while matching GPT-5.5. The model leads all open weights entries by a wide margin, outpacing MiniMax-M3 at 1408 and several proprietary systems including Gemini 3.5 Flash. Across 1,999 matches averaging 31 turns per task, GLM-5.2 also tops open weights on the Artificial Analysis Intelligence Index and Agentic Index, creating fresh pressure on GPT-5.6, Gemini 3.5Pro and next Opus releases to deliver larger gains or risk losing ground in the open source AI models race.