OpenAI
Reasoning

o3

Reasoning

Headline reasoning model in the o-series. Trades latency for depth on math, science and code benchmarks.

o3 is a reasoning AI model from OpenAI. It costs $2.00 per million input tokens and $8.00 per million output tokens (blended $3.80/M), with a 200,000-token context window.

RatesUSD per million tokens. Blend is 70% input, 30% output, our calculation.

INPUT
$2.00/M
per million input tokens
OUTPUT
$8.00/M
per million output tokens
BLENDED 70/30
$3.80/M
unchanged since 3 May
CONTEXT
200,000
tokens

Checked against OpenAI's pricing page on 10 Nov 2026.

What it is good at
  • Best-in-class on math & science benchmarks
  • Multi-step planning
  • Tool use
Typical use cases
  • Hard math/science questions
  • Complex coding
  • Research assistants

Benchmarks

vs. best public score
MMLU92.9%
Multitask academic knowledge across 57 subjects. Source: OpenAI (self-reported), MMLU, read 2026-11-10.
Graduate-level science questions, "Google-proof". Source: OpenAI (self-reported), GPQA diamond (no tools), read 2026-11-10.
HumanEval87.4%
Python function synthesis from docstrings. Source: OpenAI (self-reported), HumanEval, read 2026-11-10.
Real GitHub issues solved end-to-end. Source: OpenAI (self-reported), SWE-bench Verified, read 2026-11-10.
LMArena Elo1432 Elo
Crowd-sourced head-to-head preference Elo rating. Source: LMArena (independent), LMArena Text Arena score, read 2026-11-10.
Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources, treat as directional rather than authoritative.

Also reported by the provider

Results on benchmarks outside the set we compare across models. They are listed for completeness and are deliberately excluded from the bars above and from every ranking on this site, because a score is only comparable against the same benchmark variant. Why benchmark scores do not compare →

  • MATH-50097.8%self-reported by OpenAI · 2025-07 · o3 default effort, no tools. simple-evals README, zero-shot chain of thought. Its footnote says models from o1 onward are scored on MATH-500, not the full MATH test set. Not comparable with the MATH column, so shown here and never ranked.

More from OpenAI

See all 28 →

Frequently asked questions

How much does o3 cost?

o3 costs $2.00 per million input tokens and $8.00 per million output tokens, for a blended reference rate of $3.80 per million tokens.

What is o3's context window?

o3 supports up to 200,000 tokens of context in a single request.

What is o3 best for?

o3 is well suited to Best-in-class on math & science benchmarks, Multi-step planning and Tool use.

Who makes o3?

o3 is developed and served by OpenAI.

Compared with

Terms used on this page