Model

OpenAI o3-mini (High)

CLOSED

OpenAI · o3 family · released Jan 31, 2025

Highest reasoning-effort setting: best accuracy of the o3-mini family, higher latency/cost per query.

ReasoningCodingVisionFunction callingTool useAgentic

1622.7

Elo · rank #113

Parameters

Undisclosed

Active params

—

Context

200K tokens

Architecture

Dense transformer reasoning model (size undisclosed)

License

Proprietary

Languages

—

API price (in/out)

$1.1 / $4.4

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

AIMEMath87.3%#57

best: GPT-5.2 · 100.0%

CodeforcesCoding2130#13

best: DeepSeek-V4-Pro (Think Max) · 3206

DROPReasoning80.6#24

best: Hunyuan-T1 · 93.1

GDPval-AAAgents468#50

best: Claude Fable 5 · 1932

GPQA DiamondReasoning79.7%#89

best: GPT-6 Astra · 96.0%

Humanity's Last ExamReasoning12.0%#86

best: Claude Opus 5 · 64.7%

HumanEvalCoding97.6%#2

best: Claude Opus 4.5 · 99.4%

IFBenchReasoning67.0%#34

best: MiniMax M3 · 83.0%

LiveCodeBenchCoding67.4%#77

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MATH-500Math97.9%#10

best: GPT-5 · 99.4%

MGSMMath92.0%#4

best: OpenAI o4-mini · 93.7%

MMLUKnowledge86.9%#37

best: OpenAI o3 · 92.9%

SimpleQAKnowledge13.8%#28

best: GPT-4.5 · 62.5%

SWE-bench VerifiedCoding49.3%#94

best: Claude Opus 5 · 96.0%

Terminal-Bench 2.0Coding4.0%#77

best: Gemini 3.8 Flash · 89.4%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

$1.1

Output / M tok

$4.4

o3 family

Elo progression across releases

OpenAI o3-mini (High)Jan 20251622.7 OpenAI o3-mini (Low)Jan 20251309.0 OpenAI o3-mini (Medium)Jan 20251462.9 OpenAI o3Apr 20251896.3 OpenAI o3-proJun 20252067.0

Resources

DOCSdevelopers.openai.com/api/docs/models/o3-mini APIAvailable via OpenAI API

API price $1.1/$4.4 · each benchmark row carries its own source badge (see methodology)