Quick answers aren't always the best, which could be the key takeaway from the arrival of OpenAI Strawberry, now called o1-preview, a new ChatGPT reasoning model that takes longer to deliver what could be much better answers.
OpenAI announced the release of the preview on Thursday in a blog post, saying it's coming to ChatGPT and the generative AI company's API. I can confirm that the preview version o1 and a faster, cheaper model, the o1-mini, are available on our ChatGPT Plus account. However, the new models won't appear on free ChatGPT accounts yet.
The strawberry's appearance was eagerly awaited because of its potential ability to think in a human-like way. In the weeks leading up to this announcement, OpenAI CEO Sam Altman had teased us with numerous risqué references to fruit, but he had also made it clear in recent months that generative AI was set to take a huge leap forward.
In the blog post, OpenAI explained: “We train these models to spend more time thinking about problems before responding, similar to how a person would. Through training, they learn to refine their thinking process, try different strategies, and recognize their mistakes.”
PhD in Intelligence
OpenAI claims that this more powerful o1-preview model has performed “similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology.” And that’s key here. o1-preview is a generative model that might have the most application in academia, not in helping you write a compelling graduation proposal.
An example given in a video accompanying the blog is genetic sequencing. In it, a scientist points out that while humans can’t keep track of everything involved in genetic sequencing, an AI can. The scientist refers to the new model as “reasoning chat” and shows how, when you type in a question, there is a moment where o1-preview says “Thinking.” The value of this is that it stops you from getting into the wrong part of genetic theory.
However, o1-preview is not a replacement for ChatGPT-4o, which is barely a month old. The new model does not search the web nor is it capable of ingesting files and images. However, that will likely become noticeable at some point.
If biology and maths aren't your thing, the lighter and slightly snappier o1-mini might be just the thing for you. It's now also available in ChatGPT Plus. It's particularly well suited for coding.
You can try out the new models in ChatGPT Plus ($20/month) by logging in and selecting the model dropdown. You'll notice that o1-preview and o1-mini have been added to the list at the time of publishing this story.
It will be interesting to see what people do with the model formerly known as Strawberry and whether they get impatient as they wait for an o1 preview to deliver answers. Perhaps they'll feel better if they can see that the additional reasoning is resulting in more useful answers to some really big problems.
