Fine tuned open model ftw
retained text
another win for GLM. Chenru Duan on X: "We’re introducing Hydrogen-1.0, our first open-weight model for scientific research. We post-trained GLM-5.3 (@Zai\_org) on our internal scientific data, lab records, and discovery trajectories. It scores 38.3% on FrontierScience-Research vs GPT-6-Astra’s 33.3%
document fetched 2026-11-03 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1xm2xbq/fine_tuned_open_model_ftw