Hydrogen-1.0 open science model, post-trained on GLM-5.3, beats GPT-6-Astra on FrontierScience
Choice_Celery9481 · reddit · 2026-11-03
Chenru Duan's team introduced Hydrogen-1.0, their first open-weight model for scientific research, post-trained from GLM-5.3 on internal scientific data, lab records, and discovery trajectories.
It scores 38.3% on FrontierScience-Research, beating GPT-6-Astra's 33.3% — a notable win for open-weight models and for GLM as a base for scientific fine-tuning.
Related event: Deep Principle Open-Sources Science Model Hydrogen-1.0, Beating GPT-6-Astra(4 posts)→
More from Models
- Running Claude Opus 5.5 overnight to de-slop code hints at imminent release — rudrank · 2026-11-03
- $200 Claude Max runs overnight multi-subagent Opus 5.5 job, burns only 6% of weekly quota — lxfater · 2026-11-03
- Liquid AI's Antidoom cuts reasoning model doom loops from 10.2% to 1.4% with FTPO — maximelabonne · 2026-11-03
- Is the open-model gap real? Debate over GDPx benchmark index design — teortaxesTex · 2026-11-03
- MOSS-Live launches: full-duplex real-time voice AI that listens while it talks and runs tasks in the background — bdsqlsz · 2026-11-03
- Claude drafting a talk's speaker notes sounds hilariously like Tony Robbins — joelbot3000 · 2026-11-03