Compare · ModelsLive · 2 picked · head to head
DeepSeek V4 Flash 0731 vs Gemini 3 Pro
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
DeepSeek V4 Flash 0731 wins on 5/9 benchmarks
DeepSeek V4 Flash 0731 wins 5 of 9 shared benchmarks. Leads in reasoning · math · general.
Category leads
reasoning·DeepSeek V4 Flash 0731knowledge·Gemini 3 Promath·DeepSeek V4 Flash 0731general·DeepSeek V4 Flash 0731coding·Gemini 3 Pro
Hype vs Reality
Attention vs performance
DeepSeek V4 Flash 0731
#76 by perf·#6 by attention
Gemini 3 Pro
#71 by perf·#5 by attention
Best value
DeepSeek V4 Flash 0731
DeepSeek V4 Flash 0731
85.5 pts/$
$0.65/M
Gemini 3 Pro
n/a
no price
Vendor risk
Mixed exposure
One or more vendors flagged
DeepSeek
$3.4B·Tier 1
Google DeepMind
$4.20T·Tier 1
Head to head
9 benchmarks · 2 models
DeepSeek V4 Flash 0731Gemini 3 Pro
ARC-AGI
DeepSeek V4 Flash 0731 leads by +14.0
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
DeepSeek V4 Flash 0731
89.0
Gemini 3 Pro
75.0
ARC-AGI-2
DeepSeek V4 Flash 0731 leads by +30.3
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
DeepSeek V4 Flash 0731
61.4
Gemini 3 Pro
31.1
Chess Puzzles
DeepSeek V4 Flash 0731 leads by +2.1
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
DeepSeek V4 Flash 0731
29.5
Gemini 3 Pro
27.4
GPQA diamond
Gemini 3 Pro leads by +2.1
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
DeepSeek V4 Flash 0731
88.0
Gemini 3 Pro
90.2
OTIS Mock AIME 2024-2025
DeepSeek V4 Flash 0731 leads by +3.1
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
DeepSeek V4 Flash 0731
94.4
Gemini 3 Pro
91.4
Proofbench
DeepSeek V4 Flash 0731 leads by +36.0
DeepSeek V4 Flash 0731
56.0
Gemini 3 Pro
20.0
SimpleBench
Gemini 3 Pro leads by +18.4
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
DeepSeek V4 Flash 0731
53.3
Gemini 3 Pro
71.7
SimpleQA Verified
Gemini 3 Pro leads by +39.3
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
DeepSeek V4 Flash 0731
33.6
Gemini 3 Pro
72.9
WeirdML
Gemini 3 Pro leads by +7.0
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
DeepSeek V4 Flash 0731
63.0
Gemini 3 Pro
69.9
Full benchmark table
| Benchmark | DeepSeek V4 Flash 0731 | Gemini 3 Pro |
|---|---|---|
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 89.0 | 75.0 |
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. | 61.4 | 31.1 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 29.5 | 27.4 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 88.0 | 90.2 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 94.4 | 91.4 |
Proofbench | 56.0 | 20.0 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 53.3 | 71.7 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 33.6 | 72.9 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 63.0 | 69.9 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.02 | $1.28 | 1.0M tokens (~524 books) | $3.31 | |
| — | — | — | — |