DeepSeek Coder V2 — Benchmark Results
DeepSeek Overall Score: 58/100
Performance Scores
Overall
58
Rank #30 of 45 — Top 33%
SWE-bench
50
Rank #30 of 45 — Top 33%
LiveCodeBench
60
Rank #30 of 45 — Top 33%
HumanEval
82
Rank #27 of 45 — Top 40%
BigCodeBench
42
Rank #31 of 45 — Top 31%
Strengths & Weaknesses
Strengths
- Code-specialized
- Very cheap
Weaknesses
- Limited general-purpose
Compare with Similar-Priced Models
| Model | Overall Score | Input $/M |
|---|---|---|
| DeepSeek Coder V2 | 58 | $0.270 |
| Claude 3.5 Haiku | 52 | $0.800 |
| Claude 3 Haiku | 45 | $0.250 |
| GPT-4o mini | 58 | $0.150 |
| GPT-3.5 Turbo | 40 | $0.500 |
| OpenAI o1-mini | 70 | $1.10 |