Compare · ModelsLive · 2 picked · head to head
Gemini 3 Flash Preview vs o4 Mini (batch)
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Gemini 3 Flash Preview wins on 16/16 benchmarks
Gemini 3 Flash Preview wins 16 of 16 shared benchmarks. Leads in reasoning · knowledge · general.
Category leads
reasoning·Gemini 3 Flash Previewknowledge·Gemini 3 Flash Previewgeneral·Gemini 3 Flash Previewmath·Gemini 3 Flash Previewcoding·Gemini 3 Flash Preview
Hype vs Reality
Attention vs performance
Gemini 3 Flash Preview
#135 by perf·#10 by attention
o4 Mini (batch)
#115 by perf·no signal
Best value
o4 Mini (batch)
1.3x better value than Gemini 3 Flash Preview
Gemini 3 Flash Preview
28.5 pts/$
$1.75/M
o4 Mini (batch)
37.9 pts/$
$1.38/M
Vendor risk
Who is behind the model
Google DeepMind
$3.53T·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
16 benchmarks · 2 models
Gemini 3 Flash Previewo4 Mini (batch)
ARC-AGI
Gemini 3 Flash Preview leads by +26.0
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Gemini 3 Flash Preview
84.7
o4 Mini (batch)
58.7
ARC-AGI-2
Gemini 3 Flash Preview leads by +27.5
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
Gemini 3 Flash Preview
33.6
o4 Mini (batch)
6.1
Chess Puzzles
Gemini 3 Flash Preview leads by +20.1
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Gemini 3 Flash Preview
38.0
o4 Mini (batch)
17.9
Dtbench
Gemini 3 Flash Preview leads by +19.1
Gemini 3 Flash Preview
81.8
o4 Mini (batch)
62.7
FrontierMath-Tier-4-v2-Private
Gemini 3 Flash Preview leads by +12.2
Gemini 3 Flash Preview
17.1
o4 Mini (batch)
4.9
FrontierMath-Tiers-1-3-v2-Private
Gemini 3 Flash Preview leads by +15.9
Gemini 3 Flash Preview
53.9
o4 Mini (batch)
38.0
GeoBench
Gemini 3 Flash Preview leads by +24.0
GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding.
Gemini 3 Flash Preview
88.0
o4 Mini (batch)
64.0
GPQA diamond
Gemini 3 Flash Preview leads by +13.6
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Gemini 3 Flash Preview
89.4
o4 Mini (batch)
75.8
GSO-Bench
Gemini 3 Flash Preview leads by +6.2
GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues.
Gemini 3 Flash Preview
9.8
o4 Mini (batch)
3.6
Lmca
Gemini 3 Flash Preview leads by +19.6
Gemini 3 Flash Preview
50.7
o4 Mini (batch)
31.2
Mystery Game Puzzles
Gemini 3 Flash Preview leads by +18.5
Gemini 3 Flash Preview
18.5
o4 Mini (batch)
0.0
OTIS Mock AIME 2024-2025
Gemini 3 Flash Preview leads by +11.1
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Gemini 3 Flash Preview
92.8
o4 Mini (batch)
81.7
SimpleBench
Gemini 3 Flash Preview leads by +26.9
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
Gemini 3 Flash Preview
53.3
o4 Mini (batch)
26.4
SimpleQA Verified
Gemini 3 Flash Preview leads by +47.8
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Gemini 3 Flash Preview
67.4
o4 Mini (batch)
19.6
VPCT
Gemini 3 Flash Preview leads by +22.6
VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations.
Gemini 3 Flash Preview
58.9
o4 Mini (batch)
36.3
WeirdML
Gemini 3 Flash Preview leads by +9.0
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Gemini 3 Flash Preview
61.6
o4 Mini (batch)
52.6
Full benchmark table
| Benchmark | Gemini 3 Flash Preview | o4 Mini (batch) |
|---|---|---|
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 84.7 | 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. | 33.6 | 6.1 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 38.0 | 17.9 |
Dtbench | 81.8 | 62.7 |
FrontierMath-Tier-4-v2-Private | 17.1 | 4.9 |
FrontierMath-Tiers-1-3-v2-Private | 53.9 | 38.0 |
GeoBench GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding. | 88.0 | 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. | 89.4 | 75.8 |
GSO-Bench GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues. | 9.8 | 3.6 |
Lmca | 50.7 | 31.2 |
Mystery Game Puzzles | 18.5 | 0.0 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 92.8 | 81.7 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 53.3 | 26.4 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 67.4 | 19.6 |
VPCT VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations. | 58.9 | 36.3 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 61.6 | 52.6 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.50 | $3.00 | 1.0M tokens (~524 books) | $11.25 | |
| $0.55 | $2.20 | 200K tokens (~100 books) | $9.63 |