GPT-5.3 Codex 51.8% Accuracy on AA-Omniscience Good Enough

https://oscar-wiki.win/index.php/Why_200K_to_2M_Token_Context_Windows_Change_How_Organizations_Do_Ethical_Analysis_and_Make_Value_Judgments

OpenAI Codex Reliability and Challenges in Coding Model Hallucination Understanding GPT-5.3 Knowledge Performance on Coding Tasks As of March 2026, GPT-5.3, a key release in the lineage of OpenAI Codex models, registered roughly 51