OpenAI unveils new models designed to think more before answering

ChatGPT maker OpenAI has unveiled a new range of AI models designed to spend more time thinking before they respond, with the aim of reasoning more before helping to solve more complex tasks or problems.

Unlike its mainstream ChatGPT chatbot, the OpenAI o1 series cannot yet browse the web for information or receive image uploads.

However, OpenAI said the models had been trained to spend more time thinking through problems input into them before responding “much like a person would”, and said the models were now able to “refine their thinking process, try different strategies and recognise their mistakes”.

“These enhanced reasoning capabilities may be particularly useful if you’re tackling complex problems in science, coding, math, and similar fields,” the company said.

“For example, o1 can be used by healthcare researchers to annotate cell sequencing data, by physicists to generate complicated mathematical formulas needed for quantum optics, and by developers in all fields to build and execute multi-step workflows.”

Alongside the main version, OpenAI said it was also rolling out a “faster, cheaper” version called o1-mini which it said was “particularly effective” at coding.

“As a smaller model, o1-mini is 80% cheaper than o1-preview, making it a powerful, cost-effective model for applications that require reasoning but not broad world knowledge,” OpenAI said.

The announcement highlights how the AI firm is looking to diversify beyond just ChatGPT its accompanying tools for consumers, as it looks to capitalise on the AI frenzy.

Earlier this week, it was reported that the company was in talks with investors to raise 6.5 billion dollars (£5 billion) at a valuation of 150 billion dollars (£115 billion), making it one of the most valuable start-ups in the world.

In addition to the announcements around its new models, OpenAI also revealed that it had “formalised agreements” with AI safety institutes in the UK and US, and confirmed it had granted both institutes “early access to a research version” of the new models.

“This was an important first step in our partnership, helping to establish a process for research, evaluation, and testing of future models prior to and following their public release,” OpenAI said.