Compare · ModelsLive · 2 picked · head to head
Gemini 3 Pro vs GPT-5.4 Pro
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-5.4 Pro wins on 7/10 benchmarks
GPT-5.4 Pro wins 7 of 10 shared benchmarks. Leads in reasoning · knowledge · math.
Category leads
reasoning·GPT-5.4 Proknowledge·GPT-5.4 Promath·GPT-5.4 Procoding·Gemini 3 Pro
Hype vs Reality
Attention vs performance
Gemini 3 Pro
#52 by perf·no signal
GPT-5.4 Pro
#41 by perf·no signal
Vendor risk
Who is behind the model
Google DeepMind
$4.00T·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
10 benchmarks · 2 models
Gemini 3 ProGPT-5.4 Pro
ARC-AGI
GPT-5.4 Pro leads by +19.5
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Gemini 3 Pro
75.0
GPT-5.4 Pro
94.5
ARC-AGI-2
GPT-5.4 Pro leads by +52.2
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 Pro
31.1
GPT-5.4 Pro
83.3
Chess Puzzles
GPT-5.4 Pro leads by +27.6
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Gemini 3 Pro
31.0
GPT-5.4 Pro
58.6
FrontierMath-2025-02-28-Private
GPT-5.4 Pro leads by +12.4
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
Gemini 3 Pro
37.6
GPT-5.4 Pro
50.0
FrontierMath-Tier-4-2025-07-01-Private
GPT-5.4 Pro leads by +18.8
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
Gemini 3 Pro
18.8
GPT-5.4 Pro
37.5
GPQA diamond
GPT-5.4 Pro leads by +2.6
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Gemini 3 Pro
90.2
GPT-5.4 Pro
92.8
HLE
GPT-5.4 Pro leads by +7.1
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%.
Gemini 3 Pro
34.4
GPT-5.4 Pro
41.5
SimpleBench
Gemini 3 Pro leads by +2.8
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
Gemini 3 Pro
71.7
GPT-5.4 Pro
68.9
SimpleQA Verified
Gemini 3 Pro leads by +25.1
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Gemini 3 Pro
72.9
GPT-5.4 Pro
47.8
WeirdML
Gemini 3 Pro leads by +12.5
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Gemini 3 Pro
69.9
GPT-5.4 Pro
57.4
Full benchmark table
| Benchmark | Gemini 3 Pro | GPT-5.4 Pro |
|---|---|---|
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 75.0 | 94.5 |
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. | 31.1 | 83.3 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 31.0 | 58.6 |
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. | 37.6 | 50.0 |
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. | 18.8 | 37.5 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 90.2 | 92.8 |
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%. | 34.4 | 41.5 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 71.7 | 68.9 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 72.9 | 47.8 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 69.9 | 57.4 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| — | — | — | — | |
| $30.00 | $180.00 | 1.1M tokens (~525 books) | $675.00 |