Compare · ModelsLive · 2 picked · head to head
GPT-5 Mini vs o4 Mini
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-5 Mini wins on 19/25 benchmarks
GPT-5 Mini wins 19 of 25 shared benchmarks. Leads in knowledge · general · math.
Category leads
reasoning·o4 Miniknowledge·GPT-5 Minigeneral·GPT-5 Minimath·GPT-5 Minilanguage·o4 Minicoding·GPT-5 Mini
Hype vs Reality
Attention vs performance
GPT-5 Mini
#108 by perf·#4 by attention
o4 Mini
#132 by perf·no signal
Best value
GPT-5 Mini
2.6x better value than o4 Mini
GPT-5 Mini
45.8 pts/$
$1.13/M
o4 Mini
17.6 pts/$
$2.75/M
Vendor risk
Who is behind the model
OpenAI
$840.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
25 benchmarks · 2 models
GPT-5 Minio4 Mini
ARC-AGI
o4 Mini leads by +4.3
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
GPT-5 Mini
54.3
o4 Mini
58.7
ARC-AGI-2
o4 Mini leads by +1.7
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
GPT-5 Mini
4.4
o4 Mini
6.1
Chess Puzzles
GPT-5 Mini leads by +4.2
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
GPT-5 Mini
26.4
o4 Mini
22.1
Dtbench
GPT-5 Mini leads by +4.9
GPT-5 Mini
67.5
o4 Mini
62.7
Fiction.LiveBench
o4 Mini leads by +8.4
Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination.
GPT-5 Mini
69.4
o4 Mini
77.8
FrontierMath-2025-02-28-Private
GPT-5 Mini leads by +2.4
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
GPT-5 Mini
27.2
o4 Mini
24.8
FrontierMath-Tier-4-2025-07-01-Private
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
GPT-5 Mini
6.3
o4 Mini
6.3
FrontierMath-Tier-4-v2-Private
GPT-5 Mini leads by +7.3
GPT-5 Mini
12.2
o4 Mini
4.9
FrontierMath-Tiers-1-3-v2-Private
GPT-5 Mini leads by +10.5
GPT-5 Mini
46.7
o4 Mini
36.1
GPQA diamond
o4 Mini leads by +6.1
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
GPT-5 Mini
66.7
o4 Mini
72.8
HELM · GPQA
GPT-5 Mini leads by +2.1
GPT-5 Mini
75.6
o4 Mini
73.5
HELM · IFEval
o4 Mini leads by +0.2
GPT-5 Mini
92.7
o4 Mini
92.9
HELM · MMLU-Pro
GPT-5 Mini leads by +1.5
GPT-5 Mini
83.5
o4 Mini
82.0
HELM · Omni-MATH
GPT-5 Mini leads by +0.2
GPT-5 Mini
72.2
o4 Mini
72.0
HELM · WildBench
GPT-5 Mini leads by +0.1
GPT-5 Mini
85.5
o4 Mini
85.4
HLE
GPT-5 Mini leads by +1.4
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%.
GPT-5 Mini
15.4
o4 Mini
13.9
Lech Mazur Writing
GPT-5 Mini leads by +8.1
Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication.
GPT-5 Mini
83.1
o4 Mini
75.0
Lmca
GPT-5 Mini leads by +9.1
GPT-5 Mini
40.2
o4 Mini
31.2
MATH level 5
GPT-5 Mini leads by +0.0
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
GPT-5 Mini
97.8
o4 Mini
97.8
Mystery Game Puzzles
GPT-5 Mini leads by +0.9
GPT-5 Mini
0.9
o4 Mini
0.0
OTIS Mock AIME 2024-2025
GPT-5 Mini leads by +5.0
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
GPT-5 Mini
86.7
o4 Mini
81.7
SimpleQA Verified
GPT-5 Mini leads by +2.0
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
GPT-5 Mini
21.6
o4 Mini
19.6
SWE-Bench Verified (Bash Only)
GPT-5 Mini leads by +14.8
SWE-Bench Verified (Bash Only) · a curated subset of SWE-bench where models fix real Python repository bugs using only bash commands, no agent frameworks.
GPT-5 Mini
59.8
o4 Mini
45.0
VPCT
o4 Mini leads by +25.9
VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations.
GPT-5 Mini
10.3
o4 Mini
36.3
WeirdML
GPT-5 Mini leads by +0.1
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
GPT-5 Mini
52.7
o4 Mini
52.6
Full benchmark table
| Benchmark | GPT-5 Mini | o4 Mini |
|---|---|---|
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 54.3 | 58.7 |
ARC-AGI-2 ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data. | 4.4 | 6.1 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 26.4 | 22.1 |
Dtbench | 67.5 | 62.7 |
Fiction.LiveBench Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination. | 69.4 | 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. | 27.2 | 24.8 |
FrontierMath-Tier-4-2025-07-01-Private FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning. | 6.3 | 6.3 |
FrontierMath-Tier-4-v2-Private | 12.2 | 4.9 |
FrontierMath-Tiers-1-3-v2-Private | 46.7 | 36.1 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 66.7 | 72.8 |
HELM · GPQA | 75.6 | 73.5 |
HELM · IFEval | 92.7 | 92.9 |
HELM · MMLU-Pro | 83.5 | 82.0 |
HELM · Omni-MATH | 72.2 | 72.0 |
HELM · WildBench | 85.5 | 85.4 |
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%. | 15.4 | 13.9 |
Lech Mazur Writing Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication. | 83.1 | 75.0 |
Lmca | 40.2 | 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. | 97.8 | 97.8 |
Mystery Game Puzzles | 0.9 | 0.0 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 86.7 | 81.7 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 21.6 | 19.6 |
SWE-Bench Verified (Bash Only) SWE-Bench Verified (Bash Only) · a curated subset of SWE-bench where models fix real Python repository bugs using only bash commands, no agent frameworks. | 59.8 | 45.0 |
VPCT VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations. | 10.3 | 36.3 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 52.7 | 52.6 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.25 | $2.00 | 400K tokens (~200 books) | $6.88 | |
| $1.10 | $4.40 | 200K tokens (~100 books) | $19.25 |