rotative
the decoder · November 22, 2026

OpenAI's GPT-6.2 Astra and Sol cut answer length 34-40% but lag on transparency

openai's gpt-6.2 astra and sol produce shorter answers, get to the point faster, but admit uncertainty less often and score lower on transparency than anthropic's claude, affecting user trust and model reliability.

read in 简体中文 · 繁體中文

openai's gpt-6.2 astra and sol models have been evaluated by artificial analysis, showing significant changes in communication style. astra's average answer length decreased by 40% from 352 to 210 words, while sol's decreased by 34% from 337 to 223 words. both models use fewer headings and em dashes, and their tendency to admit uncertainty dropped sharply, with astra's score falling from 96 to 24 and sol's from 98 to 51. compared to anthropic's claude fable 5.2 and opus 5.5, astra and sol score lower on structure and transparency, with opus 5.5 posting the highest informativeness score. these changes affect user trust and model reliability, signaling a shift in openai's approach to ai communication.

read at source ↗

← rotative