Compare · ModelsLive · 2 picked · head to head
o1 vs o4 Mini
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
o4 Mini wins on 16/20 benchmarks
o4 Mini wins 16 of 20 shared benchmarks. Leads in coding · reasoning · knowledge.
Category leads
coding·o4 Minireasoning·o4 Miniknowledge·o4 Minigeneral·o4 Minimath·o4 Mini
Hype vs Reality
Attention vs performance
o1
#158 by perf·no signal
o4 Mini
#132 by perf·no signal
Best value
o4 Mini
14.6x better value than o1
o1
1.2 pts/$
$37.50/M
o4 Mini
17.6 pts/$
$2.75/M
Vendor risk
Who is behind the model
OpenAI
$840.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
20 benchmarks · 2 models
o1o4 Mini
Aider polyglot
o4 Mini leads by +10.3
Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework.
o1
61.7
o4 Mini
72.0
ARC-AGI
o4 Mini leads by +28.0
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
o1
30.7
o4 Mini
58.7
CadEval
o4 Mini leads by +6.0
CadEval · evaluates the ability to generate and reason about Computer-Aided Design code, testing spatial reasoning and engineering knowledge.
o1
56.0
o4 Mini
62.0
Chess Puzzles
o4 Mini leads by +11.6
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
o1
10.6
o4 Mini
22.1
Dtbench
o4 Mini leads by +4.9
o1
57.8
o4 Mini
62.7
Fiction.LiveBench
o1 leads by +5.5
Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination.
o1
83.3
o4 Mini
77.8
FrontierMath-2025-02-28-Private
o4 Mini leads by +15.5
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
o1
9.3
o4 Mini
24.8
FrontierMath-Tiers-1-3-v2-Private
o4 Mini leads by +21.4
o1
14.7
o4 Mini
36.1
GeoBench
o1 leads by +16.0
GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding.
o1
80.0
o4 Mini
64.0
GPQA diamond
o4 Mini leads by +3.8
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
o1
69.0
o4 Mini
72.8
HLE
o4 Mini leads by +10.6
HLE (Humanity's Last Exam) · a reasoning benchmark designed to be the hardest public evaluation of AI. Questions span mathematics, physics, philosophy, and logic · curated to be at or beyond the frontier of human expert capability. Tested with and without tool augmentation. Claude Opus 4.7 scores 46.9% without tools and 54.7% with tools · making it one of the few benchmarks where the top score is below 60%.
o1
3.3
o4 Mini
13.9
Lech Mazur Writing
o4 Mini leads by +4.8
Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication.
o1
70.2
o4 Mini
75.0
Lmca
o4 Mini leads by +5.0
o1
26.2
o4 Mini
31.2
MATH level 5
o4 Mini leads by +3.1
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
o1
94.7
o4 Mini
97.8
Metr Time Horizons
o4 Mini leads by +12.9
o1
51.0
o4 Mini
63.9
OTIS Mock AIME 2024-2025
o4 Mini leads by +8.3
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
o1
73.3
o4 Mini
81.7
SimpleBench
o1 leads by +1.7
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
o1
28.1
o4 Mini
26.4
SimpleQA Verified
o1 leads by +21.5
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
o1
41.1
o4 Mini
19.6
VPCT
o4 Mini leads by +30.8
VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations.
o1
5.5
o4 Mini
36.3
WeirdML
o4 Mini leads by +6.5
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
o1
46.1
o4 Mini
52.6
Full benchmark table
| Benchmark | o1 | o4 Mini |
|---|---|---|
Aider polyglot Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework. | 61.7 | 72.0 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 30.7 | 58.7 |
CadEval CadEval · evaluates the ability to generate and reason about Computer-Aided Design code, testing spatial reasoning and engineering knowledge. | 56.0 | 62.0 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 10.6 | 22.1 |
Dtbench | 57.8 | 62.7 |
Fiction.LiveBench Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination. | 83.3 | 77.8 |
FrontierMath-2025-02-28-Private FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning. | 9.3 | 24.8 |
FrontierMath-Tiers-1-3-v2-Private | 14.7 | 36.1 |
GeoBench GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding. | 80.0 | 64.0 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 69.0 | 72.8 |
HLE HLE (Humanity's Last Exam) · a reasoning benchmark designed to be the hardest public evaluation of AI. Questions span mathematics, physics, philosophy, and logic · curated to be at or beyond the frontier of human expert capability. Tested with and without tool augmentation. Claude Opus 4.7 scores 46.9% without tools and 54.7% with tools · making it one of the few benchmarks where the top score is below 60%. | 3.3 | 13.9 |
Lech Mazur Writing Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication. | 70.2 | 75.0 |
Lmca | 26.2 | 31.2 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 94.7 | 97.8 |
Metr Time Horizons | 51.0 | 63.9 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 73.3 | 81.7 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 28.1 | 26.4 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 41.1 | 19.6 |
VPCT VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations. | 5.5 | 36.3 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 46.1 | 52.6 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
People also compared