The new models, called o1 and o1-mini, have been officially available on ChatGPT since September 12.
The model also improved its ability to answer competitive programming questions and achieved PhD-level accuracy on a scale measuring scientific problems, according to a statement from OpenAI, Reuters reported. Noam Brown, a researcher at OpenAI, said the model achieved this feat through a technique called “chaining of thought.” The company also automated this ability so that the model could break down problems without human input.
Source: https://thanhnien.vn/openai-ra-mat-mo-hinh-ai-co-kha-nang-ly-luan-185240913231856434.htm
Comment (0)