o3
OpenAI's most powerful reasoning model, built to tackle complex problems in coding, math, science, and visual analysis with exceptional depth and accuracy.
OpenAI Deep reasoning for math, code, and vision
OpenAI o3 is the flagship model in OpenAI's o-series of reasoning models, released in April 2025. It is designed to spend more time thinking through problems before responding, using large-scale reinforcement learning to work through complex, multi-step tasks. The model supports a 200,000-token context window and can process both text and images as inputs. According to OpenAI, o3 makes 20% fewer major errors than its predecessor on difficult real-world tasks, with particular strength in programming, business consulting, and creative ideation.
A notable feature of o3 is its ability to integrate images directly into its reasoning process — not just interpreting them, but actively using them as part of problem-solving, including handling blurry, reversed, or low-quality visuals. The model can also autonomously combine tools such as web search, Python-based data analysis, and image generation to address multi-faceted questions. It is best suited for users who need rigorous analytical reasoning across domains like biology, mathematics, engineering, and software development, particularly when tasks require combining visual and textual information.
What o3 supports
Multi-Step Reasoning
Uses large-scale reinforcement learning to work through complex, multi-step problems, with performance improving the more reasoning time it is given.
Visual Reasoning
Integrates images directly into its reasoning chain, including the ability to interpret blurry, reversed, or low-quality images and manipulate visuals as part of problem-solving.
Agentic Tool Use
Autonomously combines tools such as web search, Python-based data analysis, and image generation to tackle multi-faceted questions, typically completing tasks in under a minute.
Long Context Window
Supports a 200,000-token context window, enabling processing of very long documents and complex workflows requiring large amounts of context.
Math & Science Tasks
Achieves 98.4% pass@1 on AIME 2025 math competition problems with tool access and scored 87.7% on the GPQA Diamond expert-level science benchmark.
Code Generation
Sets benchmark results on SWE-bench for software engineering and Codeforces for competitive programming, supporting complex code generation and debugging tasks.
Ready to build with o3?
Get Started FreeBenchmark scores
Scores represent accuracy — the percentage of questions answered correctly on each test.
| Benchmark | What it tests | Score |
|---|---|---|
| MMLU-Pro | Expert knowledge across 14 academic disciplines | 85.3% |
| GPQA Diamond | PhD-level science questions (biology, physics, chemistry) | 82.7% |
| MATH-500 | Undergraduate and competition-level math problems | 99.2% |
| AIME 2024 | American math olympiad problems | 90.3% |
| LiveCodeBench | Real-world coding tasks from recent competitions | 80.8% |
| HLE | Questions that challenge frontier models across many domains | 20.0% |
| SciCode | Scientific research coding and numerical methods | 41.0% |
Common questions about o3
What is the context window size for o3?
o3 supports a 200,000-token context window, which allows it to process very long documents and handle complex workflows that require large amounts of context in a single request.
What is the training data cutoff for o3?
Based on the available metadata, o3's training date is listed as April 2025. For the most precise knowledge cutoff date, refer to OpenAI's official model release notes.
Does o3 support image inputs?
Yes. o3 can accept images as inputs and incorporates them directly into its reasoning process. It can interpret blurry, reversed, or low-quality images and use visual manipulation as part of solving a problem.
What kinds of tasks is o3 best suited for?
o3 is designed for tasks requiring deep analytical reasoning, including complex coding, mathematics, scientific hypothesis evaluation, and problems that combine visual and textual information. It is particularly noted for performance in programming, business consulting, and creative ideation.
Can o3 use external tools autonomously?
Yes. o3 supports agentic tool use, meaning it can autonomously invoke tools such as web search, Python-based data analysis, and image generation to address multi-step questions, typically within under a minute.
What people think about o3
Community discussions around o3 have highlighted its long context comprehension and reasoning depth, with some users describing its performance on complex tasks as notably thorough. Early threads praised its multimodal and analytical capabilities shortly after its April 2025 release.
However, significant controversy emerged later, with a highly upvoted thread alleging that OpenAI was silently routing o3 requests to GPT-5, raising concerns about transparency and model identity. Users have also discussed broader trust issues with OpenAI's model deployment practices alongside interest in the o3-pro variant.
OpenAI's post-training lead leaves and joins Anthropic: he helped ship GPT-5, 5.1, 5.2, 5.3-Codex, o3 and o1 and will return to hands-on RL research at Anthropic
OpenAI are now stealth routing all o3 requests to GPT-5
WHAT!! OpenAI strikes back. o3 is pretty much perfect in long context comprehension.
OpenAI just pulled the biggest bait-and-switch in AI history and I'm done.
OpenAI announce o3-pro release today
Parameters & options
Used to give the model guidance on how many reasoning tokens it should generate before creating a response to the prompt. Low will favor speed and economical token usage, and high will favor more complete reasoning at the cost of more tokens generated and slower responses. The default value is medium, which is a balance between speed and reasoning accuracy.
Explore similar models
OpenAI GPT‑5.2 Pro
The most capable model series yet for professional knowledge work.
OpenAI GPT-5.1
The best model for coding and agentic tasks across domains
OpenAI GPT-5
The best model for coding and agentic tasks across domains
Google Gemini 3.1 Pro
Google's frontier reasoning model delivering enhanced software engineering, agentic reliability, and multimodal intelligence across a 1M-token context window.
Nvidia Nemotron 3 Nano 30B
NVIDIA's compact open-weight hybrid MoE model with 30B total / 3.5B active parameters, delivering strong reasoning and coding performance up to 1M context.
X.ai Grok 4.20 Reasoning
A beta reasoning variant of xAI's Grok 4.20, designed for enhanced step-by-step thinking and complex problem-solving.
Start building with o3
No API keys required. Create AI-powered workflows with o3 in minutes — free.