OpenAI o1
In plain wordsOpenAI's first reasoning model, from 2024. Before replying it writes out a long, hidden series of steps, trying ideas and checking them, which lifted its scores on hard math, science and coding problems far above earlier chatbots'.
The first of OpenAIOpenAI, prop. n.: The American company that makes ChatGPT and the GPT models. It began in 2015 as a nonprofit research lab and is now a for-profit company overseen by a nonprofit foundation.'s o-series reasoning modelsreasoning model, n.: An AI model trained to work through a problem step by step in hidden 'thinking' text before it answers. Letting it think longer generally makes it better at math, coding and planning., previewed on 12 September 2024 as o1-preview and o1-mini and released in full on 5 December 2024. Large-scale reinforcement learningreinforcement learning, n.: A way of teaching a computer program by trial and error: it tries actions, earns points for good results and penalties for bad ones, and gradually learns which choices pay off over time. taught it to work through an extended hidden chain of thoughtchain of thought, n.: Getting an AI to write out its reasoning step by step before giving an answer, the way a student shows their working. This often makes answers to math and logic problems more accurate. before giving an answer — trying approaches, checking its work and backtracking — and its accuracy on hard problems climbed as it was allowed more test-time computetest-time compute, n.: The computing effort an AI model spends while answering a question, as opposed to while being trained. Letting a model think longer or try many answers can make it markedly smarter. for thinking, opening a second axis for scaling besides training compute. In OpenAI's figures, o1 solved 83 percent of the 2024 American Invitational Mathematics Examination problems when many samples were combined by majority vote, where GPT-4oGPT-4o, prop. n.: An OpenAI model from 2024 that handles text, images and speech in one system, enabling fast, natural voice conversations. Many users grew attached to its personality, which made its retirement controversial. managed about 12 percent, and it outscored PhD-level experts on the GPQA science benchmark. OpenAI hid the raw reasoning and showed only summaries, citing competition and safety monitoring. Rival laboratories released their own reasoning models within months.
We send only the hardest tickets to o1, because its thinking tokens make it slow and expensive.
Origin
Announced in OpenAI's post 'Learning to Reason with LLMs'. During development the work was reported under the code name Strawberry, which press reports linked to an earlier OpenAI project known as Q*Q*, prop. n.: The code name of an internal OpenAI research project, reported in late 2023, said to let AI solve math problems it had never seen. Later reports linked it to OpenAI's first reasoning model, o1. (late 2023). OpenAI restarted its numbering at 1 to signal a new kind of model.
Usage note
Its successor, OpenAI o3OpenAI o3, prop. n.: OpenAI's second-generation reasoning model, announced in December 2024. Its high scores on a puzzle test designed to resist AI made headlines, and the released version could use tools such as web search while thinking., was announced three months later, and the o-series was folded into GPT-5GPT-5, prop. n.: OpenAI's flagship AI model from August 2025, which decides by itself whether to answer quickly or think longer. It replaced earlier models in ChatGPT and was followed by frequent upgrades. in August 2025. DeepSeek-R1DeepSeek-R1, prop. n.: A free, downloadable AI reasoning model released by the Chinese company DeepSeek in January 2025. It matched leading American models at math and coding for a fraction of the reported cost, shaking the industry. (January 2025) showed that similar behavior could be trained and published openly.
Related
- reasoning modelreasoning model, n.: An AI model trained to work through a problem step by step in hidden 'thinking' text before it answers. Letting it think longer generally makes it better at math, coding and planning.
- test-time computetest-time compute, n.: The computing effort an AI model spends while answering a question, as opposed to while being trained. Letting a model think longer or try many answers can make it markedly smarter.
- chain of thoughtchain of thought, n.: Getting an AI to write out its reasoning step by step before giving an answer, the way a student shows their working. This often makes answers to math and logic problems more accurate.
- OpenAI o3OpenAI o3, prop. n.: OpenAI's second-generation reasoning model, announced in December 2024. Its high scores on a puzzle test designed to resist AI made headlines, and the released version could use tools such as web search while thinking.
- DeepSeek-R1DeepSeek-R1, prop. n.: A free, downloadable AI reasoning model released by the Chinese company DeepSeek in January 2025. It matched leading American models at math and coding for a fraction of the reported cost, shaking the industry.
- GPT-5GPT-5, prop. n.: OpenAI's flagship AI model from August 2025, which decides by itself whether to answer quickly or think longer. It replaced earlier models in ChatGPT and was followed by frequent upgrades.
- reinforcement learning with verifiable rewardsreinforcement learning with verifiable rewards, n.: Training an AI by letting it attempt problems whose answers can be checked automatically — a math result, a passing program — and rewarding it only when the check succeeds.
Sources
- OpenAI (2024). Learning to Reason with LLMs. OpenAI blog, 12 September 2024.
- OpenAI (2024). OpenAI o1 System Card. December 2024.
Mentioned in
ChatGPTChatGPT, prop. n.: OpenAI's AI chatbot, released free to the public in November 2022, which answers questions, writes and converses in everyday language. Its sudden popularity began the current AI boom., GPT-5GPT-5, prop. n.: OpenAI's flagship AI model from August 2025, which decides by itself whether to answer quickly or think longer. It replaced earlier models in ChatGPT and was followed by frequent upgrades., OpenAIOpenAI, prop. n.: The American company that makes ChatGPT and the GPT models. It began in 2015 as a nonprofit research lab and is now a for-profit company overseen by a nonprofit foundation., DeepSeek momentDeepSeek moment, n.: The shock of January 2025, when a little-known Chinese company released a free AI model nearly as good as the best American ones at a fraction of the reported cost, briefly wiping hundreds of billions off chip stocks., DeepSeek-R1DeepSeek-R1, prop. n.: A free, downloadable AI reasoning model released by the Chinese company DeepSeek in January 2025. It matched leading American models at math and coding for a fraction of the reported cost, shaking the industry., RL scalingRL scaling, n.: The practice, and the finding behind it, of giving AI models far more trial-and-error practice after their initial training, especially on problems with checkable answers, which has so far improved their reasoning fairly predictably., LibratusLibratus, prop. n.: A poker-playing AI from Carnegie Mellon University that in 2017 decisively beat four top professionals at two-player no-limit Texas hold'em. Its successor, Pluribus, beat professionals at six-player poker in 2019., OpenAI o3OpenAI o3, prop. n.: OpenAI's second-generation reasoning model, announced in December 2024. Its high scores on a puzzle test designed to resist AI made headlines, and the released version could use tools such as web search while thinking., Q*Q*, prop. n.: The code name of an internal OpenAI research project, reported in late 2023, said to let AI solve math problems it had never seen. Later reports linked it to OpenAI's first reasoning model, o1.