Key Takeaways
DeepSeek V3.1 wins:
- Higher intelligence benchmark
- Better at coding
- Better at math
Qwen2.5 72B Instruct wins:
- Cheaper input tokens
- Cheaper output tokens
- Faster response time
Price Advantage
Qwen2.5 72B Instruct
Benchmark Advantage
DeepSeek V3.1
Context Window
Qwen2.5 72B Instruct
Speed
Qwen2.5 72B Instruct
Pricing Comparison
Price Comparison
| Metric | DeepSeek V3.1 | Qwen2.5 72B Instruct | Winner |
|---|---|---|---|
| Input (per 1M tokens) | $0.15 | $0.12 | Qwen2.5 72B Instruct |
| Output (per 1M tokens) | $0.75 | $0.39 | Qwen2.5 72B Instruct |
Using a 3:1 input/output ratio, Qwen2.5 72B Instruct is 38% cheaper overall.
DeepSeek V3.1 Providers
SambaNova $0.15 (Cheapest)
Chutes $0.20
DeepInfra $0.21
SiliconFlow $0.27
Novita $0.27
Qwen2.5 72B Instruct Providers
DeepInfra $0.12 (Cheapest)
Chutes $0.30
Novita $0.38
Hyperbolic $0.40
Benchmark Comparison
8
Benchmarks Compared
6
DeepSeek V3.1 Wins
0
Qwen2.5 72B Instruct Wins
Benchmark Scores
| Benchmark | DeepSeek V3.1 | Qwen2.5 72B Instruct | Winner |
|---|---|---|---|
Intelligence Index Overall intelligence score | 28.0 | 15.6 | |
Coding Index Code generation & understanding | 28.4 | 11.9 | |
Math Index Mathematical reasoning | 49.7 | 14.0 | |
MMLU Pro Academic knowledge | 83.3 | 72.0 | |
GPQA Graduate-level science | 73.5 | 49.1 | |
LiveCodeBench Competitive programming | 57.7 | 27.6 | |
Aider Real-world code editing | - | 65.4 | - |
AIME Competition math | - | 16.0 | - |
DeepSeek V3.1 significantly outperforms in coding benchmarks.
Cost vs Quality
X-axis:
Y-axis:
Loading chart...
DeepSeek V3.1
Other models
Context & Performance
Context Window
DeepSeek V3.1
32,768
tokens
Max output: 7,168 tokens
Qwen2.5 72B Instruct
32,768
tokens
Max output: 16,384 tokens
Speed Performance
| Metric | DeepSeek V3.1 | Qwen2.5 72B Instruct | Winner |
|---|---|---|---|
| Tokens/second | 0.0 tok/s | 45.6 tok/s | |
| Time to First Token | 0.00s | 1.11s |
Qwen2.5 72B Instruct responds Infinity% faster on average.
Capabilities
Feature Comparison
| Feature | DeepSeek V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| Vision (Image Input) | ||
| Tool/Function Calls | ||
| Reasoning Mode | ||
| Audio Input | ||
| Audio Output | ||
| PDF Input | ||
| Prompt Caching | ||
| Web Search |
License & Release
| Property | DeepSeek V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| License | Open Source | Open Source |
| Author | Deepseek | Qwen |
| Released | Aug 2025 | Sep 2024 |
DeepSeek V3.1 Modalities
Input
text
Output
text
Qwen2.5 72B Instruct Modalities
Input
text
Output
text