OpenAI logo

o3

OpenAI · Released Apr 24, 2025

ProprietaryreasoningcodingCompare this model

$2

Input price / MTok

#18 of 27

$8

Output price / MTok

#17 of 27

200K

Context window

#24 of 28

167

Tokens / sec

#4 of 21

71.7

SWE-bench Verified

#9 of 12

87.7

GPQA Diamond

#10 of 19

88.9

AIME 2025

#5 of 10

82.9

MMMU

#2 of 8

Summary

o3 is a proprietary language model from OpenAI, released on Apr 24, 2025.

At $2 input and $8 output per million tokens, it ranks #18 of 27 for input price, placing it mid-range on input cost.

Its context window of 200K tokens ranks #24 of 28 for context depth in this dataset.

It targets reasoning, coding workloads.

Throughput is measured at 167 tokens per second (ranked #4 of 21 for throughput).

Full reasoning model; chain-of-thought reasoning tokens billed as output.

Benchmarks

BenchmarkScoreMax%RankSource
SWE-bench Verified71.710071.7%#9/12OpenAI (self-reported at o3/o4-mini laun...
GPQA Diamond87.710087.7%#10/19OpenAI (self-reported at o3/o4-mini laun...
AIME 202588.910088.9%#5/10OpenAI (self-reported at o3/o4-mini laun...
MMMU82.910082.9%#2/8OpenAI (self-reported at o3/o4-mini laun...

Similar Models

Compare all 6 models

Models with the most shared benchmark coverage and closest scores to o3.

o3
OpenAIGPT-5
GoogleGemini 2.5 Pro
OpenAIo4-mini
AnthropicClaude Sonnet 4.6
AnthropicClaude Haiku 4.5
SWE-bench Verified#9 of 12
Real-world GitHub issue resolution
GPQA Diamond#10 of 19
Graduate-level science reasoning
AIME 2025#5 of 10
Advanced math competition problems
MMMU#2 of 8
Multimodal understanding across 30 subjects

Compare o3 side by side

Pre-filled with 5 similar models

Open comparison

Speed source: https://artificialanalysis.ai/models/o3