Z.ai CEO Jie Tang: GLM 5.3 gains come from RL, not parameter count

Z.ai CEO Jie Tang argues parameter count is only meaningful alongside data volume, compute strategy, and deployment conditions. GLM-5.3's gains come solely from RL on long-horizon production environments, some spanning days of engineering work; memorization prefers parameters while reasoning prefers post-training data.

Featured · Jie Tang