↓Skip to main content
Zeitgeist — a spike by Chris Gathercole
  1. Creators/
  2. Simon Willison — Creator of Datasette; co-creator of Django/

Quoting Ted Sanders

Blog · Watch/Read · (gathered 2026-10-29)

  • Ted Sanders explains that LLM performance is incredibly multidimensional, but much of its variation is primarily explained by model size, with bigger models generally being more performant, reliable, and costly.
  • He notes that bigger models, such as OpenAI’s ordered progression (Luna < Terra < Sol < Astra), also tend to have better recall of esoteric facts.
  • Sanders clarifies that OpenAI’s version numbers (e.g., 6.1 vs. 5.7 vs. 7) signify not just release order, but also the model’s relationship to other releases (e.g., shared “personality”) and the magnitude of its improvement.