Toby Ord: Newer AI Models Show Improving Scaling Slope γ, Not Just Curve Shifts
tobyordoxford · x · 2026-10-09
Philosopher Toby Ord extends the AISI scaling charts analysis: γ can be read off directly as the slope of each model's scaling curve.
- Model improvements typically appear as sideways shifts of the curve, with the slope unchanged
- Interestingly, more recent models have shown better γ values — the slope itself is improving
- He plots γ as a function of model release date and compares γ values with the token scale-up needed to match a 10x increase in human task time
This suggests the newest models may be gaining qualitative, not merely incremental, improvements in time-horizon scaling.
Related event: UK AISI Data Shows Model Time Horizons Scale Sublinearly with Tokens(4 posts)→
More from Models
- Test shows Haiku 5.5 triples robot task success with more thinking, GPT-6 Luna stays near zero — ycombinator · 2026-10-10
- Open-EmbeddingGemma thread wraps up: why Kye Gomez reimplements closed-source models — KyeGomezB · 2026-10-10
- Open-EmbeddingGemma: pure PyTorch reimplementation of Google's 740M multimodal embedding model — KyeGomezB · 2026-10-10
- Sources behind the Open-EmbeddingGemma reimplementation: HF model page and Google blog — KyeGomezB · 2026-10-10
- How EmbeddingGemma 2 works: six steps from five modalities to one 768-dim vector — KyeGomezB · 2026-10-10
- Dev claims Gemini 3.8 Flash beats Opus 5.5 for traditional RAG in finance — rickasaurus · 2026-10-10