Friday, March 13, 2026

OpenAI's new GPT-5.4 clobbers humans on pro-level work in tests - by 83% - David Gewirtz, ZDnet

GPT-5.4 is also more reliable, producing 18% fewer errors and 33% fewer false claims than GPT-5.2, according to OpenAI. GPT-5.4's 83% score suggests AI rivals expert professionals. Tests span nine industries and 44 real-world occupations. New capabilities boost coding, tools, and computer control.