Get the latest tech news

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


GPT-5.4 is also more reliable, producing 18% fewer errors and 33% fewer false claims than GPT-5.2, according to OpenAI.

None

Get the Android app

Or read this on ZDNet

Read more on:

Photo of OpenAI

OpenAI

Photo of tests

tests

Photo of Humans

Humans

Related news:

News photo

OpenAI Releases New Financial-Services Tools, Rivaling Anthropic

News photo

OpenAI launches GPT-5.4 with Pro and Thinking versions

News photo

Anthropic CEO Dario Amodei Calls OpenAI's Messaging Around Military Deal 'Straight Up Lies'