Unverified: GPT-6 Astra matches GLM-5.3 max reasoning with zero reasoning tokens

BLUECOW009 · x · 2026-09-06

An unverified community claim: GPT-6 Astra with reasoning turned off is on par with GLM-5.3 max reasoning, completing the entire Artificial Analysis index with 7 million tokens and zero reasoning tokens. If true, the base model alone matches a competitor's full reasoning tier. No official confirmation yet.

Original post →

More from Models

Models channel →