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.
More from Models
- GPT 6.2 Astra hits PostTrainBench: slightly beats 6 Astra, still trails Opus 5.5 — maksym_andr · 2026-11-21
- Live Verification of Claude's Riemann Proof Hits 99%; Kontorovich Weighs In — AlexKontorovich · 2026-11-21
- GPT 6.2 Astra lands on PostTrainBench, still trailing Anthropic's Opus 5.5 — maksym_andr · 2026-11-21
- Google brings Gemini 3.5 Transcribe voice editing to the Gemini API — GeminiApp · 2026-11-21
- Have frontier models found an exploit in LEAN? Reddit weighs testable signs — wabawanga · 2026-11-21
- Kapitsa runs Hy4-flash on M5 Max at 36 tok/s decode, 971 tok/s prefill at 32K — m4rkGS · 2026-11-21