Researcher Calls Early GPT-6 Results Underwhelming

Rylan Schaeffer, a Stanford CS PhD student on leave joining Meta, posted on X that he ran one of his tests on early GPT-6. He reported the results as underwhelming, with the same mistakes appearing as in GPT 5.6 Sol. Schaeffer stated the improvement did not match the noticeable increase seen from 5.4 to 5.5. The post reflects his assessment of the model version he evaluated. No other details or confirmations appear in the packet.