Bindu Reddy stated that while ZAI's GLM-5.2 model performs exceptionally well on benchmarks, it s..., Sonic AI
“Bindu Reddy stated that while ZAI's GLM-5.2 model performs exceptionally well on benchmarks, it shows signs of being over-tuned for them, with internal evaluations placing it behind Opus 4.8 and GPT-5.5.”