GPT-6 Astra reportedly tops FrontierSWE at 65.5%, nearly doubling predecessor

charliermarsh · x · 2026-09-18

ProximalHQ claims GPT-6 Astra tops the FrontierSWE benchmark with 65.5%, beating Fable 5.1 (56.3%) and its predecessor GPT-5.6 Sol (32.2%). The claim is unverified by OpenAI and should be treated with caution.

Related event: Report: GPT-6 Astra Tops FrontierSWE Benchmark(2 posts)→

Original post →

More from Models

Models channel →