GPT-6 Astra Debuts at No.3 on Creative Writing Benchmark, Blind Tests Favor Itself

GPT-6 Astra (high) debuted at No.3 on Lech Mazur's Creative Writing Benchmark, beating GPT-5.6 Sol, though blind self-evaluation revealed a bias toward its own outputs, choosing itself in 692 of 700 rounds.

2026-09-06 ~ 2026-09-06 · 2 related posts

• GPT-6 Astra hits #3 on creative writing benchmark; blind self-judging exposes 692/700 self-bias — zero0_one1 · 2026-09-06

1 near-duplicate retellings: morqon