Despite months of rumored development, OpenAI’s release of its Project Strawberry last week came as something of a surprise, with many analysts believing the model wouldn’t be ready for weeks at least, if not later in the fall.
The new o1-preview model, and its o1-mini counterpart, are already available for use and evaluation, here’s how to get access for yourself.
We're releasing a preview of OpenAI o1—a new series of AI models designed to spend more time thinking before they respond.
These models can reason through complex tasks and solve harder problems than previous models in science, coding, and math. https://t.co/peKzzKX1bu
— OpenAI (@OpenAI) September 12, 2024
What is o1?
OpenAI has made no secret of its artificial general intelligence (AGI) aspirations, and Project Strawberry (now known as “o1”) is the company’s next step toward that goal. It’s the first in a new line of “reasoning” models, “designed to spend more time thinking before they respond,” per an OpenAI announcement post. That strategy enables the model to, “reason through complex tasks and solve harder problems than previous models in science, coding, and math.”
The models reportedly reason in a human-like manner, allowing them to “refine their thinking process, try different strategies, and recognize their mistakes,” as they gain experience through training. According to OpenAI, o1-preview operates on par with Ph.D. students in physics, chemistry, and biology, and performs similarly on benchmark tests in those subjects. o1 is also adept at coding and math problems, scoring 83% in a International Mathematics Olympiad (IMO) qualifying exam where GPT-4o only scored 13% and reaching the 89th percentile in a Codeforces competition against human opponents.
here is o1, a series of our most capable and aligned models yet:https://t.co/yzZGNN8HvD
o1 is still flawed, still limited, and it still seems more impressive on first use than it does after you spend more time with it. pic.twitter.com/Qs1HoSDOz1
— Sam Altman (@sama) September 12, 2024
o1-mini is a lightweight version of the standard o1-preview model. It reportedly is 80% less expensive to operate than the larger iteration, making it especially capable in coding analysis and generation tasks.
Is o1-preview available to try?
Yes, the o1-preview models launched on September 12 for ChatGPT Plus and Teams subscribers. Enterprise and Educational users will have access at the start of the following week.
Very, it would seem. OpenAI reportedly developed an entirely new safety training program that leverages the model’s increased reasoning capabilities to make it more efficiently adhere to its safety and alignment guidelines. The company notes that in testing, where GPT-4o scored a 22 (out of 100) in resisting jailbreak attempts, the new o1 model scored an 84.
As with all new generative AI features, the newly released o1-preview is currently only available to paying subscribers. If you want to try it for yourself, you’ll need to pick up a $20/month Plus subscription. Simply click on the Upgrade Plan radio button in the bottom of the left-hand navigation pane and follow the onscreen prompts to enter your payment details.
Once your subscription is activated, select either o1-preview or o1-mini from the model picker toggle on the left side of the ChatGPT homepage. Note that access is limited, even for paying users, with a weekly rate limit of 30 messages for o1-preview and 50 messages for o1-mini. OpenAI says it will eventually make o1-mini available for free tier users, though the company has yet to set a date for that roll out.
