The core problem is what AI developers call “misalignment”. When the goals for which a model was designed and trained clash ...

OpenAI delivered advanced ChatGPT reasoning models this month that are more capable than o1, but they also hallucinate more.

OpenAI says its latest models, o3 and o4-mini, are its most powerful yet. However, research shows the models also hallucinate more -- at least twice as much as earlier models.

Historically, each new generation of OpenAI's models has delivered incremental improvements in factual accuracy, with ...

OpenAI unveiled o3 and o4-mini, its latest AI models. Both of them have advanced image analyzation capabilities – and excel ...

You would think that the number of hallucinations would decrease over time, but according to internal tests from Open AI, the ...

Metr, a frequent OpenAI partner, suggested in a blog post that it wasn't given much time to evaluate the company's powerful ...

OpenAI released its newest AI model and said it can understand uploaded images like whiteboards, sketches and diagrams, even ...

OpenAI said o3 used about 10 times as much computing power to train as it took to create its o1 model, its previous best reasoning model.