o1
OpenAI · September 2024
● activeCloseddecoder onlytextAPI Available
Context Window200K tokens
Variantso1, o1-mini, o1-preview
Why It Matters
Pioneered the 'thinking model' paradigm, proving that letting AI reason longer at inference time dramatically improves performance on hard problems. Opened an entirely new scaling dimension beyond just training bigger models.
Description
OpenAI's first dedicated reasoning model. Instead of answering immediately, it 'thinks' step-by-step before responding — a technique called test-time compute, where the model spends extra processing time during inference to work through complex problems. Excels at math, coding, and scientific reasoning that stumps conventional models.
Notable Milestones
- ▸Achieved expert-level performance on PhD-level science questions
- ▸Scored in the 89th percentile on competitive programming (Codeforces)
- ▸Ranked in the top 500 students in the USA Math Olympiad qualifier
Benchmark Scores
GPQAGraduate-level science QA
78.0%AIMEAMC/AIME math competition
83.3%SWE-benchReal-world software engineering
48.9%Key Innovations
Reasoning
ReasoningStructured step-by-step problem solving, often using chain-of-thought or tree-of-thought approaches.
Chain-of-Thought
Chain-of-ThoughtPrompting technique where the model 'thinks out loud' step by step before giving a final answer.
Test-Time Compute
Test-Time ComputeUsing extra computation during inference (not training) to improve answer quality — thinking longer on harder problems.
Family Tree
Built On
Lineage
Successors (1)
Related Research (1)
Chain-of-ThoughtReasoning
2022 · Google
Showed that prompting models to "think step-by-step" unlocks arithmetic, logic, and commonsense reasoning in large models like PaLM.