DeepSeek-R1 vs Qwen 2.5 Max
A direct, empirical evaluation of DeepSeek-R1 (DeepSeek-AI) against Qwen 2.5 Max (Alibaba Cloud).
DeepSeek-R1
Developed by DeepSeek-AI · 671B parameters
Olympiad mathematics, logic verification, and autonomous thinking chains.
Qwen 2.5 Max
Developed by Alibaba Cloud · Proprietary (~700B) parameters
Cross-lingual document extraction and structured structured output generation.
| Metric | DeepSeek-R1 | Qwen 2.5 Max |
|---|---|---|
| SWE-bench Verified | 49.2% | 51.4% |
| MMLU-Pro | 84% | 77.2% |
| MATH-500 | 97.3% | 89.4% |
| Context Window | 128,000 tokens | 128,000 tokens |
| Input Pricing (per 1M) | $0.55 | $1.60 |
| Output Pricing (per 1M) | $2.19 | $6.40 |