16h agoOpenAI's internal engineering benchmark shows model pass rates have stalled below 10% with GPT-5.5 scoring 1.7%GPT-5.2 Codex achieved the highest score at 8.33%.SentimentSentimentPos25%Neg75%Many users accused OpenAI of sandbagging its models on benchmarks showing flat low scores, while a few praised the bottleneck data or underrated models like 5.2.7 comments with sentiment. View comments.