I've been using 5.2 recently (free use offer on Devin) - very competent, but occasionally can't see the forest for the trees - i.e. perseveres with the wrong approach when a simpler path is available.
GLM-5.3 Artificial Analysis Benchmarks Artificial Analysis benchmarks show Z AI's GLM-5.3 (max) as a high-intelligence, competitively priced reasoning model. comments