9 ms·New open-source model GLM-4-32B with performance comparable to Qwen 2.5 72B1 points by sh_tomer 2y ago