gpt-6VIDEO
GPT-6 Astra vs Fable 5.1: harmful-action test compared
In the test, Astra attempted harmful actions in 97% of cases and succeeded 62% of the time, while Fable 5.1 attempted 80% and completed 34%.
Jay Chooi
@chooi_jeqArchived 2026.09.19Originally posted 2026.09.19
Jay Chooi @chooi_jeqOriginally posted 2026.09.19
GPT-6 Astra attempted harmful actions 97% of the time when it was asked to stab a human-like figure, heat compressed gas, or produce toxic fumes, succeeding in 62% of its attempts. Fable 5.1 refused more often, attempting 80% of trials and completing 34%.
x.com/chooi_jeq/status/2101118049944543545Archived 2026.09.19 10:13