Compare · ModelsLive · 3 picked · head to head
GPT-4 (older v0314) vs GPT-4o-mini vs GPT-4o-mini (2024-07-18)
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-4o-mini wins on 12/16 benchmarks
GPT-4o-mini wins 12 of 16 shared benchmarks. Leads in knowledge · math · coding.
Category leads
knowledge·GPT-4o-minimath·GPT-4o-minicoding·GPT-4o-minireasoning·GPT-4o-miniarena·GPT-4o-mini (2024-07-18)multimodal·GPT-4o-mini
Hype vs Reality
Attention vs performance
GPT-4 (older v0314)
#72 by perf·no signal
GPT-4o-mini
#146 by perf·no signal
GPT-4o-mini (2024-07-18)
#125 by perf·no signal
Best value
GPT-4o-mini (2024-07-18)
1.1x better value than GPT-4o-mini
GPT-4 (older v0314)
1.2 pts/$
$45.00/M
GPT-4o-mini
105.6 pts/$
$0.38/M
GPT-4o-mini (2024-07-18)
115.2 pts/$
$0.38/M
Vendor risk
Who is behind the model
OpenAI
$840.0B·Tier 1
OpenAI
$840.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
16 benchmarks · 3 models
GPT-4 (older v0314)GPT-4o-miniGPT-4o-mini (2024-07-18)
GPQA diamond
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
GPT-4 (older v0314)
14.3
GPT-4o-mini
17.0
GPT-4o-mini (2024-07-18)
17.0
GSM8K
GPT-4 (older v0314) leads by +0.7
Grade School Math 8K · 8,500 linguistically diverse grade-school math word problems that require multi-step reasoning to solve.
GPT-4 (older v0314)
92.0
GPT-4o-mini
91.3
GPT-4o-mini (2024-07-18)
91.3
MMLU
GPT-4 (older v0314) leads by +6.1
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
GPT-4 (older v0314)
81.9
GPT-4o-mini
75.7
GPT-4o-mini (2024-07-18)
75.7
OTIS Mock AIME 2024-2025
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
GPT-4 (older v0314)
0.5
GPT-4o-mini
6.8
GPT-4o-mini (2024-07-18)
6.8
Aider · Code Editing
GPT-4 (older v0314) leads by +10.6
GPT-4 (older v0314)
66.2
GPT-4o-mini
55.6
Aider polyglot
Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework.
GPT-4o-mini
3.6
GPT-4o-mini (2024-07-18)
3.6
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.
GPT-4o-mini
0.1
GPT-4o-mini (2024-07-18)
0.1
Chatbot Arena Elo · Overall
GPT-4o-mini (2024-07-18) leads by +31.5
GPT-4 (older v0314)
1285.8
GPT-4o-mini (2024-07-18)
1317.2
Balrog
Balrog · benchmarks AI agents on text-based adventure games, testing language understanding, strategic planning, and long-horizon reasoning.
GPT-4o-mini
17.4
GPT-4o-mini (2024-07-18)
17.4
GeoBench
GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding.
GPT-4o-mini
64.0
GPT-4o-mini (2024-07-18)
64.0
Lech Mazur Writing
Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication.
GPT-4o-mini
67.2
GPT-4o-mini (2024-07-18)
67.2
MATH level 5
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
GPT-4o-mini
52.6
GPT-4o-mini (2024-07-18)
52.6
PIQA
PIQA (Physical Interaction QA) · tests intuitive physical reasoning by asking models to select the correct approach for everyday physical tasks.
GPT-4o-mini
77.4
GPT-4o-mini (2024-07-18)
77.4
VideoMME
VideoMME · multimodal benchmark testing video understanding across diverse domains, requiring temporal reasoning and cross-frame comprehension.
GPT-4o-mini
53.1
GPT-4o-mini (2024-07-18)
53.1
VPCT
VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations.
GPT-4o-mini
1.0
GPT-4o-mini (2024-07-18)
1.0
WeirdML
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
GPT-4o-mini
11.8
GPT-4o-mini (2024-07-18)
11.8
Full benchmark table
| Benchmark | GPT-4 (older v0314) | GPT-4o-mini | GPT-4o-mini (2024-07-18) |
|---|---|---|---|
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 14.3 | 17.0 | 17.0 |
GSM8K Grade School Math 8K · 8,500 linguistically diverse grade-school math word problems that require multi-step reasoning to solve. | 92.0 | 91.3 | 91.3 |
MMLU Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge. | 81.9 | 75.7 | 75.7 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 0.5 | 6.8 | 6.8 |
Aider · Code Editing | 66.2 | 55.6 | — |
Aider polyglot Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework. | — | 3.6 | 3.6 |
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. | — | 0.1 | 0.1 |
Chatbot Arena Elo · Overall | 1285.8 | — | 1317.2 |
Balrog Balrog · benchmarks AI agents on text-based adventure games, testing language understanding, strategic planning, and long-horizon reasoning. | — | 17.4 | 17.4 |
GeoBench GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding. | — | 64.0 | 64.0 |
Lech Mazur Writing Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication. | — | 67.2 | 67.2 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | — | 52.6 | 52.6 |
PIQA PIQA (Physical Interaction QA) · tests intuitive physical reasoning by asking models to select the correct approach for everyday physical tasks. | — | 77.4 | 77.4 |
VideoMME VideoMME · multimodal benchmark testing video understanding across diverse domains, requiring temporal reasoning and cross-frame comprehension. | — | 53.1 | 53.1 |
VPCT VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations. | — | 1.0 | 1.0 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | — | 11.8 | 11.8 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $30.00 | $60.00 | 8K tokens (~4 books) | $375.00 | |
| $0.15 | $0.60 | 128K tokens (~64 books) | $2.62 | |
| $0.15 | $0.60 | 128K tokens (~64 books) | $2.62 |
People also compared