Compare · ModelsLive · 2 picked · head to head
DeepSeek V4 Pro 0813 vs Gemini 3.8 Flash
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Gemini 3.8 Flash wins on 19/21 benchmarks
Gemini 3.8 Flash wins 19 of 21 shared benchmarks. Leads in speed · agentic · reasoning.
Category leads
speed·Gemini 3.8 Flashagentic·Gemini 3.8 Flashreasoning·Gemini 3.8 Flashknowledge·Gemini 3.8 Flashgeneral·Gemini 3.8 Flashcoding·Gemini 3.8 Flashmath·Gemini 3.8 Flash
Hype vs Reality
Attention vs performance
DeepSeek V4 Pro 0813
#50 by perf·#3 by attention
Gemini 3.8 Flash
#53 by perf·#11 by attention
Best value
DeepSeek V4 Pro 0813
1.9x better value than Gemini 3.8 Flash
DeepSeek V4 Pro 0813
45.5 pts/$
$1.32/M
Gemini 3.8 Flash
24.1 pts/$
$2.48/M
Vendor risk
Mixed exposure
One or more vendors flagged
DeepSeek
$3.4B·Tier 1
Google DeepMind
$4.09T·Tier 1
Head to head
21 benchmarks · 2 models
DeepSeek V4 Pro 0813Gemini 3.8 Flash
Artificial Analysis · CritPt
Gemini 3.8 Flash leads by +18.3
DeepSeek V4 Pro 0813
0.0
Gemini 3.8 Flash
18.3
Artificial Analysis · GDPval
Gemini 3.8 Flash leads by +18.9
DeepSeek V4 Pro 0813
27.9
Gemini 3.8 Flash
46.8
Artificial Analysis · Humanity's Last Exam
Gemini 3.8 Flash leads by +37.2
DeepSeek V4 Pro 0813
10.6
Gemini 3.8 Flash
47.8
Artificial Analysis · Long Context Reasoning
Gemini 3.8 Flash leads by +30.6
DeepSeek V4 Pro 0813
50.7
Gemini 3.8 Flash
81.3
Artificial Analysis · Quality Index
Gemini 3.8 Flash leads by +20.6
DeepSeek V4 Pro 0813
20.4
Gemini 3.8 Flash
40.9
Artificial Analysis · SciCode
Gemini 3.8 Flash leads by +16.7
DeepSeek V4 Pro 0813
39.9
Gemini 3.8 Flash
56.6
APEX-Agents
Gemini 3.8 Flash leads by +17.0
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
DeepSeek V4 Pro 0813
47.3
Gemini 3.8 Flash
64.3
ARC-AGI
Gemini 3.8 Flash leads by +8.0
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
DeepSeek V4 Pro 0813
90.5
Gemini 3.8 Flash
98.5
ARC-AGI-2
Gemini 3.8 Flash leads by +27.9
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 Pro 0813
61.3
Gemini 3.8 Flash
89.2
Chess Puzzles
Gemini 3.8 Flash leads by +12.6
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
DeepSeek V4 Pro 0813
43.2
Gemini 3.8 Flash
55.8
Dtbench
Gemini 3.8 Flash leads by +3.1
DeepSeek V4 Pro 0813
89.8
Gemini 3.8 Flash
92.9
Frontiercode
Gemini 3.8 Flash leads by +12.6
DeepSeek V4 Pro 0813
28.6
Gemini 3.8 Flash
41.2
FrontierMath-Tier-4-v2-Private
DeepSeek V4 Pro 0813 leads by +4.9
DeepSeek V4 Pro 0813
26.8
Gemini 3.8 Flash
21.9
FrontierMath-Tiers-1-3-v2-Private
Gemini 3.8 Flash leads by +3.9
DeepSeek V4 Pro 0813
64.6
Gemini 3.8 Flash
68.4
GPQA diamond
Gemini 3.8 Flash leads by +5.0
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
DeepSeek V4 Pro 0813
88.9
Gemini 3.8 Flash
93.9
Lmca
Gemini 3.8 Flash leads by +8.8
DeepSeek V4 Pro 0813
53.5
Gemini 3.8 Flash
62.2
Mystery Game Puzzles
Gemini 3.8 Flash leads by +4.4
DeepSeek V4 Pro 0813
37.2
Gemini 3.8 Flash
41.6
OTIS Mock AIME 2024-2025
Gemini 3.8 Flash leads by +0.3
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
DeepSeek V4 Pro 0813
98.6
Gemini 3.8 Flash
98.9
Proofbench
DeepSeek V4 Pro 0813 leads by +2.0
DeepSeek V4 Pro 0813
50.0
Gemini 3.8 Flash
48.0
SimpleQA Verified
Gemini 3.8 Flash leads by +16.8
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
DeepSeek V4 Pro 0813
52.9
Gemini 3.8 Flash
69.7
WeirdML
Gemini 3.8 Flash leads by +18.6
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
DeepSeek V4 Pro 0813
66.2
Gemini 3.8 Flash
84.8
Full benchmark table
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 3.8 Flash |
|---|---|---|
Artificial Analysis · CritPt | 0.0 | 18.3 |
Artificial Analysis · GDPval | 27.9 | 46.8 |
Artificial Analysis · Humanity's Last Exam | 10.6 | 47.8 |
Artificial Analysis · Long Context Reasoning | 50.7 | 81.3 |
Artificial Analysis · Quality Index | 20.4 | 40.9 |
Artificial Analysis · SciCode | 39.9 | 56.6 |
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | 47.3 | 64.3 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 90.5 | 98.5 |
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. | 61.3 | 89.2 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 43.2 | 55.8 |
Dtbench | 89.8 | 92.9 |
Frontiercode | 28.6 | 41.2 |
FrontierMath-Tier-4-v2-Private | 26.8 | 21.9 |
FrontierMath-Tiers-1-3-v2-Private | 64.6 | 68.4 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 88.9 | 93.9 |
Lmca | 53.5 | 62.2 |
Mystery Game Puzzles | 37.2 | 41.6 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 98.6 | 98.9 |
Proofbench | 50.0 | 48.0 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 52.9 | 69.7 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 66.2 | 84.8 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.66 | $1.98 | 1.0M tokens (~524 books) | $9.90 | |
| $0.82 | $4.13 | 1.0M tokens (~524 books) | $16.50 |