OpenAI's O1 Model: A Breakthrough in AI Reasoning

Why OpenAI's new model is such a big deal

Welcome back to The Algorithm! This week we're going to talk about OpenAI's impressive new reasoning model, called o1. I want to illustrate why it's such a big deal with an example from my wedding (bear with me -- I promise you'll see the relevance). Last weekend, I got married at a summer camp, and during the day our guests competed in a series of games inspired by the show Survivor that my now-wife and I orchestrated. When we were planning the games in August, we wanted one station to be a memory challenge, where our friends and family would have to memorize part of a poem and then relay it to their teammates so they could re-create it with a set of wooden tiles. I thought OpenAI's GPT-4o, its leading model at the time, would be perfectly suited to help. I asked it to create a short wedding-themed poem, with the constraint that each letter could only appear a certain number of times so we could make sure teams would be able to reproduce it with the provided set of tiles. GPT-4o failed miserably. The model repeatedly insisted that its poem worked within the constraints, even though it didn't. It would correctly count the letters only after the fact, while continuing to deliver poems that didn't fit the prompt. Without the time to meticulously craft the verses by hand, we ditched the poem idea and instead challenged guests to memorize a series of shapes made from colored tiles. (That ended up being a total hit with our friends and family, who also competed in dodgeball, egg tosses, and capture the flag.) However, last week OpenAI released a new model called o1 (previously referred to under the code name "Strawberry" and, before that, Q*) that blows GPT-4o out of the water for this type of purpose. Unlike previous models that are well suited for language tasks like writing and editing, OpenAI o1 is focused on multistep "reasoning," the type of process required for advanced mathematics, coding, or other STEM-based questions. It uses a "chain of thought" technique, according to OpenAI. "It learns to recognize and correct its mistakes. It learns to break down tricky steps into simpler ones. It learns to try a different approach when the current one isn't working," the company wrote in a blog post on its website. OpenAI's tests point to resounding success. The model ranks in the 89th percentile on questions from the competitive coding organization Codeforces and would be among the top 500 high school students in the USA Math Olympiad, which covers geometry, number theory, and other math topics. The model is also trained to answer PhD-level questions in subjects ranging from astrophysics to organic chemistry. In math olympiad questions, the new model is 83.3% accurate, versus 13.4% for GPT-4o. In the PhD-level questions, it averaged 78% accuracy, compared with 69.7% from human experts and 56.1% from GPT-4o. (In light of these accomplishments, it's unsurprising the new model was pretty good at writing a poem for our nuptial games, though still not perfect; it used more Ts and Ss than instructed to.) So why does this matter? The bulk of LLM progress until now has been language-driven, resulting in chatbots or voice assistants that can interpret, analyze, and generate words. But in addition to getting lots of facts wrong, such LLMs have failed to demonstrate the types of skills required to solve important problems in fields like drug discovery, materials science, coding, or physics. OpenAI's o1 is one of the first signs that LLMs might soon become genuinely helpful companions to human researchers in these fields. It's a big deal because it brings "chain-of-thought" reasoning in an AI model to a mass audience, says Matt Welsh, an AI researcher and founder of the LLM startup Fixie.

OpenAI's o1 takes a leap with model that reasons like us

OpenAI's latest model o1, dubbed as the strawberry project capable of "thinking" and "reasoning", is better suited for complex applications like software programming, STEM, legal, disease diagnosis, scientific research etc, executives and experts said. It is six times costlier than the GPT 4o Mini. "It is probably more oriented for developers that are doing math, coding, STEM, academic research versus your general tasks," said Krish Ramineni, founder of the transcription platform Firelies.ai. "So maybe someone that's building an AI software engineer, this will be more relevant. Or someone that's building an AI physics tutor, it would be more relevant." "OpenAI is betting on the fact that enterprises will increasingly use AI for reasoning in domains such as quantitative computations and software engineering," said Arun Chandrasekaran, Distinguished VP Analyst, Gartner. "It is also betting on the fact that higher latency and higher costs would be a fair trade-off for better reasoning performance," he added. While announcing the model this week, OpenAI claimed that in a qualifying exam for the International Mathematics Olympiad, GPT-4o correctly solved only 13% of problems, while the reasoning model o1 scored 83%. When tested with IIT-JEE mains questions, o1 was the only model that performed accurately, said Amit Bansal, chief product officer at ed-tech company Infinity Learn. "Strawberry AI's ability to handle complex math problems and give customised feedback shows how these models can directly improve learning," he said. "We prompted it to reveal its line of reasoning and found the chain of thought reasoning to be superior to the reasoning of GPT 4o." With advanced reasoning capabilities, the model demonstrates greater precision and reduces instances of hallucination, said Mayank Kumar, co-founder and managing director of upGrad. He explained that Reinforced Learning combined with the Chain of Thoughts method will significantly reduce energy consumption at the inference level. "Also, as we see a shift towards smaller models, these are poised to be deployed more extensively due to their efficiency," Kumar added. A large language model is no longer a tool to predict the next word or the sentence, said Pawan Prabhat, co-founder of Shorthills AI, "but it really grasps the meaning of the question, evaluates all the possible answers and gives the best one." "This new model will find extensive use in tasks that involve logic and mathematical reasoning - like software programming, legal advice, prognosis and diagnosis of diseases and scientific research," Prabhat added. Monica Malhotra Kandhari, managing director of Aasoka, an edtech platform for K-12 students, said, "For us, blending technology with education, Strawberry AI enhances our ability to create smarter, more inclusive learning environments that cater to each student's unique needs."

Why OpenAI's Reasoning Model Is Special

OpenAI finally released its Strawberry reasoning artificial intelligence last week -- or rather, an initial, less-complete version known as o1-preview. We first reported about the breakthrough behind Strawberry 10 months ago when it was still called Q*, and more recently told you what was coming, though we expected a more inspiring name than o1-preview! The reasoning model differs from prior large language models like GPT-4 in one key way: When training the reasoning model, its capabilities grow at a higher rate the more computing power you give it, thanks to the way it makes sense of, or "thinks," about data it has already reviewed. In essence, it creates new data, or thoughts, without needing as much information as prior models did.