OpenAI has released several new artificial intelligence models, beginning with the most recent OpenAI o1 model, which the company claims can “think before it answers.”
“OpenAI stated in a blog post on September 12 that this is a substantial advancement and represents a new level of AI capability for complex reasoning tasks.” “In light of this, we are resetting the counter to one and renaming this series OpenAI o1.”
OpenAI stated that the new models can take their time to consider and employ “chain-of-thought” reasoning to solve complex tasks, particularly those related to STEM (science, technology, engineering, and math) and coding.
The AI firm shared videos of OpenAI o1 solving a complex logical conundrum and coding a video game from a prompt, among other things.
The OpenAI o1 “preview” and “mini” models were made available to ChatGPT Plus subscribers, and the company intends to release enhanced versions in the near future.
OpenAI disclosed data that indicates that OpenAI o1 surpasses GPT-4o in numerous benchmarks, including certain high school examinations in the United States and PhD-level science topics in Biology, Chemistry, and Physics.
According to OpenAI, the o1 mini’s concentration on STEM reasoning capabilities renders it less proficient in subjects beyond its limited scope.
“[Its] factual knowledge on non-STEM topics, such as dates, biographies, and trivia, is comparable to small LLMs like the GPT-4o mini…”
“In addition to experimenting with extending the model to other modalities and specialties outside of STEM, we will enhance these limitations in future versions.”
Industry experts predicted that OpenAI would introduce a reasoning-focused AI model under the pseudonym Strawberry in September.
Nevertheless, OpenAI does not provide information regarding the distinctions between various models that are currently in development.
