Neon: 1,300 H200s and automated labs beat GPT-6 Astra on materials benchmark
andrew_n_carr · x · 2026-09-16
Liam Fedus describes Menlo Park high-throughput materials labs creating an experiment-model loop: labs generate fresh data, models learn and pick next experiments. With only 1,300 H200s plus months of proprietary data, mid-training and RL produced open-source Neon, surpassing GPT-6 Astra on their analysis benchmark. Andrew Carr quipped AGI is starting to feel like a harness problem.
More from Models
- 'You Can Feel the Gradient Slopes': Dev Contrasts Benchmaxxed RL Models vs General Ones — theteknosaur · 2026-11-02
- LightOnOCR-3 Launches: Tops FRBench, Adds Layout Structure and Chart Extraction — IgorCarron · 2026-11-02
- NVIDIA's Kumo Is a Family of Four Tabular Foundation Models, Not Just One — pandeyparul · 2026-11-02
- SemiAnalysis: safety classifiers are underrated — Meta's Muse tops cyber benchmark at 35.2% — ThomasScialom · 2026-11-02
- DeepSeek V4.1 Flash 1017 wins over a longtime closed-source-only user: fast, cheap, reliable — _bachrc · 2026-11-02
- Local quants vs Claude on 9 real private bugs: 3-bit costs thinking time, not score — FeydRowan · 2026-11-02