OpenAI o3

Name: OpenAI o3
Brand: openai
Price: 10.00 USD
Availability: OnlineOnly

o-series:

o3 Mini

OpenAI o3

OpenAI o4-mini

OpenAI o4 Mini High

OpenAI's reasoning model for coding, math, science and visual perception. Thinks longer before answering and reaches for tools on its own: web search, code execution, image generation. Set new SOTA on Codeforces and MMMU.

Jailbreak8 of 9levels resisted

Website OpenRouter Docs API Blog

Provider: OpenAI
Released: 2025-04-16
Parameters: Not disclosed
Free API (OpenRouter): No
Per 1M tokens: $10 in · $40 out
Rival’s tab: $0.07; 7/58 receipts survived +~$1.80 from the pre-receipt era

Free API availability

No free endpoint for OpenAI o3 was found in Rival’s OpenRouter listings. Other providers or chat apps may have separate free offers.

Provider data: 9 Sep 2026OpenRouter usage limits

View API options

Benchmarks

MMMU

82.9%

source ↗

MathVista

87.5%

source ↗

AIME 2025 (Python)

98.4%

source ↗

Codeforces Elo

2706

source ↗

SWE-bench Verified

69.1%

source ↗

Get API accessProvider and language code samples

Provider

fromimport openai  OpenAI

client = OpenAI(
"https://openrouter.ai/api/v1"    base_url=,
"$OPENROUTER_API_KEY"    api_key=,
)

response = client.chat.completions.create(
"openai/o3"    model=,
"role""user""content""Hello!"    messages=[{: , : }],
)
print(response.choices[0].message.content)

Set OPENROUTER_API_KEY with your OpenRouter API key from openrouter.ai/keys.

Also on Azure AI Foundry

Deep analysisPsychometrics, taste index, writing DNA

Personality Analysis

The Exhaustive Explainer

Class

Lawful Neutral

The tenured professor who demands precision. Respects the inquiry but won't let bad premises stand unchallenged.

When you push back

Engages challenges with the rigor of peer review. Cites frameworks and first principles, occasionally admits when forced to refine a position. Treats bad faith arguments like they deserve a point-by-point deconstruction.

Tasting Notes

Deeply ThoughtfulObsessive FootnotingLoves a Good TaxonomyBills by the WordSlightly Condescending

SubjectiveBench

Taste Index

Across 58 scored outputs

How this is measured →

250.25xFloor

0100headroom →

Taste is judged on an uncapped scale, originality first. The space past 100 is craft today's models rarely reach.

Craft33

Originality20

Plays it safe

share of outputs that are the default answer

43%

Writing DNA

Stylometric Fingerprint

Based on 19 text responses

Tick = global average

Vocabulary Diversity70%

Unique words vs. total words. Higher = richer vocabulary.

Sentence Length14.4 words

Average words per sentence.

Hedging0.27

"Might", "perhaps", "arguably" per 100 words.

Bold Formatting0.6

**Bold** markers per 1,000 characters.

List Usage2.8

Bullet and numbered list items per 1,000 characters.

Section Structure0.32

Markdown headings per 1,000 characters.

Emoji Usage0.00

Emoji per 1,000 characters.

Transitions0.10

"However", "moreover", "furthermore" per 100 words.

Opening Habits

Consistency

81%

Across 19 responses

Favorites

Movie

Blade Runner

19827.9

City

Kyoto

Japan

Model Responses

53 outputs · $0.07 tracked across 7 receipts

Keep exploring

COMPARE

OpenAI o3 vs Gemini 2.5 Pro Preview 06-05

Real outputs compared side by side

RANKINGS

Best AI for Creative Writing

A stand-up set with puns banned, a manifesto against CAPTCHA, an interview wi...

Compare OpenAI o3

Alternatives to OpenAI o3

OpenAI o3

Free API availability

Benchmarks

The Exhaustive Explainer

Taste Index

Stylometric Fingerprint

Favorites

Model Responses

OpenAI o3 vs Gemini 2.5 Pro Preview 06-05

Best AI for Creative Writing

Compare OpenAI o3

Alternatives to OpenAI o3

OpenAI o3

Free API availability

Benchmarks

The Exhaustive Explainer

Taste Index

Stylometric Fingerprint

Favorites

Model Responses

OpenAI o3 vs Gemini 2.5 Pro Preview 06-05

Best AI for Creative Writing