Compare · ModelsLive · 2 picked · head to head
DeepSeek V3 vs Qwen2.5 72B Instruct
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
DeepSeek V3 wins on 15/15 benchmarks
DeepSeek V3 wins 15 of 15 shared benchmarks. Leads in knowledge · arena · reasoning.
Category leads
knowledge·DeepSeek V3arena·DeepSeek V3reasoning·DeepSeek V3general·DeepSeek V3math·DeepSeek V3coding·DeepSeek V3
Hype vs Reality
Attention vs performance
DeepSeek V3
#79 by perf·no signal
Qwen2.5 72B Instruct
#141 by perf·#1 by attention
Best value
Qwen2.5 72B Instruct
1.4x better value than DeepSeek V3
DeepSeek V3
92.9 pts/$
$0.60/M
Qwen2.5 72B Instruct
129.7 pts/$
$0.38/M
Vendor risk
Mixed exposure
One or more vendors flagged
DeepSeek
$3.4B·Tier 1
Alibaba (Qwen)
$293.0B·Tier 1
Head to head
15 benchmarks · 2 models
DeepSeek V3Qwen2.5 72B Instruct
ARC AI2
DeepSeek V3 leads by +1.1
AI2 Reasoning Challenge · tests grade-school level science knowledge with multiple-choice questions requiring reasoning beyond simple retrieval.
DeepSeek V3
93.7
Qwen2.5 72B Instruct
92.7
Chatbot Arena Elo · Overall
DeepSeek V3 leads by +55.6
DeepSeek V3
1358.4
Qwen2.5 72B Instruct
1302.8
BBH
DeepSeek V3 leads by +10.3
BIG-Bench Hard · a curated subset of 23 challenging tasks from BIG-Bench where language models previously failed to outperform average humans.
DeepSeek V3
83.3
Qwen2.5 72B Instruct
73.1
Dtbench
DeepSeek V3 leads by +3.1
DeepSeek V3
41.3
Qwen2.5 72B Instruct
38.2
GPQA diamond
DeepSeek V3 leads by +10.3
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
Qwen2.5 72B Instruct
33.5
HellaSwag
DeepSeek V3 leads by +5.5
HellaSwag · tests commonsense reasoning by asking models to predict the most plausible continuation of everyday scenarios.
DeepSeek V3
85.2
Qwen2.5 72B Instruct
79.7
Lmca
DeepSeek V3 leads by +2.4
DeepSeek V3
18.2
Qwen2.5 72B Instruct
15.8
MATH level 5
DeepSeek V3 leads by +1.7
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
Qwen2.5 72B Instruct
63.2
Metr Time Horizons
DeepSeek V3 leads by +11.6
DeepSeek V3
47.4
Qwen2.5 72B Instruct
35.8
MMLU
DeepSeek V3 leads by +2.5
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
DeepSeek V3
82.9
Qwen2.5 72B Instruct
80.4
OTIS Mock AIME 2024-2025
DeepSeek V3 leads by +7.8
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
DeepSeek V3
15.8
Qwen2.5 72B Instruct
8.0
PIQA
DeepSeek V3 leads by +4.2
PIQA (Physical Interaction QA) · tests intuitive physical reasoning by asking models to select the correct approach for everyday physical tasks.
DeepSeek V3
69.4
Qwen2.5 72B Instruct
65.2
TriviaQA
DeepSeek V3 leads by +11.0
TriviaQA · reading comprehension benchmark with trivia questions, requiring models to find and reason over evidence from provided documents.
DeepSeek V3
82.9
Qwen2.5 72B Instruct
71.9
WeirdML
DeepSeek V3 leads by +20.1
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
DeepSeek V3
36.1
Qwen2.5 72B Instruct
16.0
Winogrande
DeepSeek V3 leads by +5.8
WinoGrande · large-scale commonsense reasoning benchmark where models must resolve ambiguous pronouns in carefully constructed sentence pairs.
DeepSeek V3
70.4
Qwen2.5 72B Instruct
64.6
Full benchmark table
| Benchmark | DeepSeek V3 | Qwen2.5 72B Instruct |
|---|---|---|
ARC AI2 AI2 Reasoning Challenge · tests grade-school level science knowledge with multiple-choice questions requiring reasoning beyond simple retrieval. | 93.7 | 92.7 |
Chatbot Arena Elo · Overall | 1358.4 | 1302.8 |
BBH BIG-Bench Hard · a curated subset of 23 challenging tasks from BIG-Bench where language models previously failed to outperform average humans. | 83.3 | 73.1 |
Dtbench | 41.3 | 38.2 |
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 | 33.5 |
HellaSwag HellaSwag · tests commonsense reasoning by asking models to predict the most plausible continuation of everyday scenarios. | 85.2 | 79.7 |
Lmca | 18.2 | 15.8 |
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 | 63.2 |
Metr Time Horizons | 47.4 | 35.8 |
MMLU Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge. | 82.9 | 80.4 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 15.8 | 8.0 |
PIQA PIQA (Physical Interaction QA) · tests intuitive physical reasoning by asking models to select the correct approach for everyday physical tasks. | 69.4 | 65.2 |
TriviaQA TriviaQA · reading comprehension benchmark with trivia questions, requiring models to find and reason over evidence from provided documents. | 82.9 | 71.9 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 36.1 | 16.0 |
Winogrande WinoGrande · large-scale commonsense reasoning benchmark where models must resolve ambiguous pronouns in carefully constructed sentence pairs. | 70.4 | 64.6 |
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 | |
| $0.36 | $0.40 | 33K tokens (~16 books) | $3.70 |
People also compared