Anthropic-trained researcher: Opus and GPT have been neck-and-neck for a year, Opus edges intent

adam_karvonen · x · 2026-11-21

Adam Karvonen argues the two labs' flagships have been effectively tied for the past year: in Nov 2025, Opus 4.5 and GPT 5.2 were equally smart with Opus slightly better at understanding intent; by Nov 2026, Opus 5.5 and GPT 6.2 show the same pattern. His interlocutor kuza55 counters that Anthropic was actually behind on model quality for most of its existence and its recent acceleration is mostly catch-up, though 5.5 was a breakout.

Related event: Researchers Question Anthropic's Claimed 2x R&D Speedup as Gap Over OpenAI Fails to Materialize(6 posts)→

Original post →

More from Models

Models channel →