LLMcompare

Search

Search for a command to run...

A · closed · GPT-4 / o-series

o1

OpenAI

B · closed · GPT-4 / o-series

o3

OpenAI

0:12

overall capability wins (2 ties across 14 scored). o3 leads overall.

  • Reasoning0:8
  • Coding0:3
  • Arena0:1
  • Pricing0:3
  • Speed0:2

Verdict

o3 leads coding (widest gap: +19.6 on Aider); o3 ranks higher on LMArena (+30 Elo); o3 is ~7.5x cheaper on a blended token basis than o1.

  • o3 leads coding (widest gap: +19.6 on Aider)
  • o3 ranks higher on LMArena (+30 Elo)
  • o3 is ~7.5x cheaper on a blended token basis than o1

Cost for 1M in + 250K out

Illustrative chat workload at primary-provider list prices.

o1
$30
o3
$4

o3 is about 7.5x cheaper on this workload.

Price per 1M tokens

Full comparison

Organization

OpenAI

OpenAI

Family

GPT-4 / o-series

GPT-4 / o-series

License

Proprietary

Proprietary

Open weights

No

No

Release

Dec 5, 2024

Apr 16, 2025

Knowledge cutoff

-

-

API / provider

OpenAI

OpenAI

Modalities

text, image → text

text, image → text

Specs

Context window

200K

200K

tie

Max output

100K

100K

tie

Parameters

—

—

—

Pricing

Input $/1M

$15

$2

B +$13

Output $/1M

$60

$8

B +$52

Blended $/1M (3∶1)

$26.25

$3.50

B +$22.75

Speed

tok/s

35

55

B +20

TTFT (s)

3

1.2

B +1.80

Reasoning

MMLU-Pro

84.1

85.3

B +1.2 pts

GPQA Diamond

74.7

82.7

B +8.0 pts

Humanity's Last Exam

7.0

20.1

B +13.1 pts

AIME 2025

—

88.9

—

MATH-500

97.0

99.2

B +2.2 pts

Humanity's Last Exam (with tools)

—

—

—

AA-Omniscience Accuracy

34.5

38.6

B +4.1 pts

AA-LCR v1.1

65.0

74.7

B +9.7 pts

CritPt

0.3

1.1

B +0.8 pts

MMMU-Pro

—

70.1

—

IFBench

70.3

71.4

B +1.1 pts

Chartography

—

—

—

Chartography (With Tools)

—

—

—

Coding

SWE-bench Verified

—

58.4

—

SWE-bench Pro

—

—

—

SWE-bench Multilingual

—

—

—

LiveCodeBench

67.9

80.8

B +12.9 pts

Terminal-Bench 2.1

—

—

—

Aider Polyglot

61.7

81.3

B +19.6 pts

Terminal-Bench 3

—

—

—

BigCodeBench

—

—

—

SciCode

35.8

41.0

B +5.2 pts

CursorBench

—

—

—

SWE-Rebench

—

38.1

—

NL2Repo-Bench

—

—

—

DeepSWE

—

—

—

WebDev Arena

—

—

—

Terminal-Bench 4.0

—

—

—

CursorBench 4.0

—

—

—

FrontierCode v1.1 (Main)

—

—

—

Arena

LMArena Elo

1,402

1,432

B +30 Elo

o1

Context
200K / 100K out
Parameters
—
Price
$15 / $60
Speed
35 tok/s · 3s TTFT
Modalities
text, image → text
License
Proprietary

Original o-series reasoning model - legacy deep reasoning SKU retained for API compatibility.

o3

Context
200K / 100K out
Parameters
—
Price
$2 / $8
Speed
55 tok/s · 1.2s TTFT
Modalities
text, image → text
License
Proprietary

OpenAI reasoning model from the o-series; strong on STEM and agentic coding with deliberate chain-of-thought.

Benchmark charts

Winner bars are emphasized. Per-benchmark deltas sit above each chart.

Reasoning & knowledge

MMLU-Pro: B +1.2GPQA: B +8.0HLE: B +13.1MATH-500: B +2.2CritPt: B +0.8AA-LCR: B +9.7Omniscience Acc.: B +4.1IFBench: B +1.1

Coding

LiveCode: B +12.9Aider: B +19.6SciCode: B +5.2

Arena

Arena Elo: B +30

Pick a different pair · Back to catalog