XNews

GPT-6.2 Astra and Sol Show Shorter, Less Structured Responses

GPT-6.2 Astra and Sol Show Shorter, Less Structured Responses

GPT-6.2 Astra and GPT-6.2 Sol changed noticeably from prior releases, giving shorter responses and using less structure. Astra's average response length fell from 352 to 210 words and Sol's from 337 to 223, while the Organization Index dropped from 70 to 11 and 86 to 27 respectively. Both models also became less likely to admit uncertainty and more natural sounding.

Event version 1.

Key insights

Shorter responses

Average response length fell from 352 to 210 words for Astra and from 337 to 223 words for Sol.

Less structured, more answer-first

The Organization Index dropped from 70 to 11 for Astra and from 86 to 27 for Sol, while Answer First rose from 49 to 89 and from 64 to 87 respectively.

Lower uncertainty admission

The Admits Uncertainty measure fell from 96 to 24 for Astra and from 98 to 51 for Sol.

More natural sounding

Natural Sounding rose from 58 to 100 for Astra and from 54 to 79 for Sol.

Evidence-based rises with effort

Evidence-Based scores rise from low to max effort, from 12 to 54 for Astra and from 16 to 68 for Sol.

Source posts