Why OpenAI and Anthropic keep swapping places on reasoning evals
teortaxesTex · x · 2026-10-28
- The author argues OpenAI and Anthropic can rapidly trade places on measurable consumer reasoning evals (math, vision, tool use, coding) because both face almost no resource constraints for RL, with mature data acquisition and RL pipelines that let them easily "countermog" each other.
- Real divergences come from AGI and product strategy: Claudes are much stronger on enterprise work and knowledge (driven by product strategy; Anthropic's pretraining lead remains a mystery to him), while GPTs are more token-efficient and stronger on frontier STEM via RLVR. Claudes also show distinctly better theory of mind.
- A follow-up note: an SWE-first small model fell off hard on science/knowledge—scale-sensitive capabilities—so it needs a stronger base model to recover.
More from Models
- Claim: Anthropic Holds Back Stronger Claude Models While OpenAI Struggles — haider1 · 2026-10-28
- LiquidAI's edge-focused 2B vision-language model LFM3-VL-2B trends on Hugging Face — LiquidAI · 2026-10-28
- Dev blogger slams OpenAI: constant reset notices, slower and dumber models — lxfater · 2026-10-28
- Datalab lite OCR costs $0.50 per 1,000 pages, 10x-30x cheaper than Azure, Google, AWS — VikParuchuri · 2026-10-28
- Screenshot Suggests Gemini 4 Argon Is Rolling Out to Pro Subscribers with Weekly Quota — arg-developer · 2026-10-28
- Gemini Notebook reportedly adding Connectors, with Google Drive as the only launch integration — testingcatalog · 2026-10-28