Toby Ord flags weak inference scaling: math gains still logarithmic at 10,000 agents
tobyordoxford · x · 2026-09-14
Toby Ord examines the inference-scaling chart OpenAI showed in its Navier–Stokes post: gains remain logarithmic, and with 10,000 concurrent agents the curve likely tracks agent count rather than inference depth. By the measure of 'fraction of maths problems solved from a big list', scaling still looks poor.
Related event: OpenAI Data Shows Reasoning Scaling Still Logarithmic(2 posts)→
More from Models
- Swift-Qwen3.8-27b, a token-efficient reasoning Qwen finetune, trends on Hugging Face — ukisai · 2026-09-14
- Thinking intensity does boost capability: R1-Zero paper cited amid DeepSeek max-vs-high setting debate — karminski3 · 2026-09-14
- GPT-6 Astra hands-on: composes first, orchestrates later, and reportedly outshines Fable and Sol — paw_lean · 2026-09-14
- Researcher posts proof he both synthesized viruses and trained a 375B open-weight LLM — ethanCaballero · 2026-09-14
- Toby Ord: 10x more RLVR compute cuts tokens-to-target ~3x; gains may be math-specific — tobyordoxford · 2026-09-14
- New scaling curve has half the slope: 10,000x compute for 20%-to-80%, but bigger generational jumps — tobyordoxford · 2026-09-14