GPT-5.5 ships: tops Terminal-Bench at 82.7%, kills Codex (again), 20% pricier

OpenAI launched GPT-5.5, scoring 82.7% on Terminal-Bench 2.0 and 73.1% on Expert-SWE, with reduced hallucinations and lower token usage at matching latency. Codex is folded back into the main model, API pricing rises 20%, and OpenAI is telling developers to throw out old prompts and rebuild from minimal baselines. ZDNET scored it 93/100 across a 10-round test, knocking points only for 'exuberance.'