GPT-6.2 Astra and Sol get to the point faster but score lower on transparency than Claude

Impact 7 The Decoder

Summary

  • An Artificial Analysis evaluation shows OpenAI's GPT-6.2 Astra and Sol models produce answers 34 to 40 percent shorter than their predecessors.
  • Astra and Sol get to the point faster and use fewer headings and em dashes, but admit uncertainty far less often.
  • Anthropic's Claude Fable 5.2 and Opus 5.5 score significantly higher on transparency, organization, and informativeness than OpenAI's latest models.
  • The personality indexes measure communication styles and relative trait rankings using AI judges, rather than overall model accuracy.

Why it matters

Evaluating communication style shifts in frontier models helps developers and architects choose models that best fit user experience requirements, contrasting OpenAI's concise responses with Anthropic's structured transparency.

Who should care

manager 60