Well ur thesis was never good and had a huge hole in it U never tested those models and checked the performance Zhipu/Knowledge Atlas has consistently released great coding models since GLM 4.6/4.7 (last year) That's not the case with Minimax M2/2.5 was good, but they have highly underperformed with M1/M2.7/M3 M3 underperformance was a major narrative loss Meanwhile, GLM is seeing larger adoption with every new release They are also expanding margins They adopted DSA for GLM5, which reduced their inference costs by a lot, but they didn't reduce API prices GLM5.2 (latest release) is a top 5 coding model globally available at a very low price