Compare · ModelsLive · 2 picked · head to head
DeepSeek V3 vs GPT-4o (2024-11-20)
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
DeepSeek V3 wins on 14/17 benchmarks
DeepSeek V3 wins 14 of 17 shared benchmarks. Leads in coding · general · math.
Category leads
coding·DeepSeek V3general·DeepSeek V3math·DeepSeek V3knowledge·DeepSeek V3language·DeepSeek V3reasoning·DeepSeek V3
Hype vs Reality
Attention vs performance
DeepSeek V3
#80 by perf·no signal
GPT-4o (2024-11-20)
#255 by perf·no signal
Best value
DeepSeek V3
17.9x better value than GPT-4o (2024-11-20)
DeepSeek V3
92.9 pts/$
$0.60/M
GPT-4o (2024-11-20)
5.2 pts/$
$6.25/M
Vendor risk
Mixed exposure
One or more vendors flagged
DeepSeek
$3.4B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
17 benchmarks · 2 models
DeepSeek V3GPT-4o (2024-11-20)
Aider polyglot
DeepSeek V3 leads by +25.3
Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework.
DeepSeek V3
48.4
GPT-4o (2024-11-20)
23.1
Dtbench
DeepSeek V3 leads by +0.4
DeepSeek V3
41.3
GPT-4o (2024-11-20)
40.9
FrontierMath-2025-02-28-Private
DeepSeek V3 leads by +2.7
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
DeepSeek V3
3.0
GPT-4o (2024-11-20)
0.3
GPQA diamond
DeepSeek V3 leads by +13.3
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
DeepSeek V3
43.8
GPT-4o (2024-11-20)
30.5
HELM · GPQA
DeepSeek V3 leads by +1.8
DeepSeek V3
53.8
GPT-4o (2024-11-20)
52.0
HELM · IFEval
DeepSeek V3 leads by +1.5
DeepSeek V3
83.2
GPT-4o (2024-11-20)
81.7
HELM · MMLU-Pro
DeepSeek V3 leads by +1.0
DeepSeek V3
72.3
GPT-4o (2024-11-20)
71.3
HELM · Omni-MATH
DeepSeek V3 leads by +11.0
DeepSeek V3
40.3
GPT-4o (2024-11-20)
29.3
HELM · WildBench
DeepSeek V3 leads by +0.3
DeepSeek V3
83.1
GPT-4o (2024-11-20)
82.8
Lech Mazur Writing
GPT-4o (2024-11-20) leads by +4.8
Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication.
DeepSeek V3
77.0
GPT-4o (2024-11-20)
81.8
Lmca
GPT-4o (2024-11-20) leads by +1.3
DeepSeek V3
18.2
GPT-4o (2024-11-20)
19.5
MATH level 5
DeepSeek V3 leads by +15.1
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
DeepSeek V3
64.8
GPT-4o (2024-11-20)
49.8
Metr Time Horizons
DeepSeek V3 leads by +6.6
DeepSeek V3
47.4
GPT-4o (2024-11-20)
40.8
MMLU
GPT-4o (2024-11-20) leads by +1.2
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
DeepSeek V3
82.9
GPT-4o (2024-11-20)
84.1
OTIS Mock AIME 2024-2025
DeepSeek V3 leads by +9.6
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
DeepSeek V3
15.8
GPT-4o (2024-11-20)
6.2
SimpleBench
DeepSeek V3 leads by +1.3
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
DeepSeek V3
2.7
GPT-4o (2024-11-20)
1.4
WeirdML
DeepSeek V3 leads by +11.0
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
DeepSeek V3
36.1
GPT-4o (2024-11-20)
25.1
Full benchmark table
| Benchmark | DeepSeek V3 | GPT-4o (2024-11-20) |
|---|---|---|
Aider polyglot Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework. | 48.4 | 23.1 |
Dtbench | 41.3 | 40.9 |
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. | 3.0 | 0.3 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 43.8 | 30.5 |
HELM · GPQA | 53.8 | 52.0 |
HELM · IFEval | 83.2 | 81.7 |
HELM · MMLU-Pro | 72.3 | 71.3 |
HELM · Omni-MATH | 40.3 | 29.3 |
HELM · WildBench | 83.1 | 82.8 |
Lech Mazur Writing Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication. | 77.0 | 81.8 |
Lmca | 18.2 | 19.5 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 64.8 | 49.8 |
Metr Time Horizons | 47.4 | 40.8 |
MMLU Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge. | 82.9 | 84.1 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 15.8 | 6.2 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 2.7 | 1.4 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 36.1 | 25.1 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.32 | $0.89 | 164K tokens (~82 books) | $4.63 | |
| $2.50 | $10.00 | 128K tokens (~64 books) | $43.75 |
People also compared