Tech Meridian ← LIVE FEED
PROMY MERIDIAN RU

FUNDING · MODELS · #4034

Artificial Analysis: GPT-6.2 Astra and Sol give much shorter answers but score lower on transparency than Anthropic

An Artificial Analysis evaluation using the AA-PersonalityBench finds OpenAI's GPT-6.2 Astra and Sol produce answers 34–40% shorter than their predecessors, use far fewer headings, and prioritize answering first, while showing large drops on the 'Admits Uncertainty' and Organization metrics (Astra: admits uncertainty 96→24; Sol: 98→51). Compared with Anthropic's Claude Fable 5.2 and Opus 5.5, Astra and Sol score lower on the Transparency and Organization indexes, while Opus 5.5 posts the highest Informativeness score.

KEY POINTS

  1. An Artificial Analysis evaluation using the AA-PersonalityBench finds OpenAI's GPT-6.2 Astra and Sol produce answers 34–40% shorter than their predecessors, use far fewer headings, and prioritize answering first, while showing large drops on the 'Admits Uncertainty' and Organization metrics (Astra: admits uncertainty 96→24; Sol: 98→51).
  2. Compared with Anthropic's Claude Fable 5.2 and Opus 5.5, Astra and Sol score lower on the Transparency and Organization indexes, while Opus 5.5 posts the highest Informativeness score.
  3. The finding matters because it shows a notable shift in OpenAI's communication style that trades more concise, direct answers for reduced explicit uncertainty signaling, which affects user trust, safety assessments, and model selection.

WHY IT MATTERS

The finding matters because it shows a notable shift in OpenAI's communication style that trades more concise, direct answers for reduced explicit uncertainty signaling, which affects user trust, safety assessments, and model selection.

SOURCES & TIMELINE

1