DeepSeek-R1 vs GPT-6.1 Sol
A direct, empirical evaluation of DeepSeek-R1 (DeepSeek-AI) against GPT-6.1 Sol (OpenAI).
DeepSeek-R1
Developed by DeepSeek-AI · 671B parameters
Olympiad mathematics, logic verification, and autonomous thinking chains.
GPT-6.1 Sol
Developed by OpenAI · Proprietary Frontier Workhorse Engine parameters
Near-Astra reasoning at 1/5th price ($1.25/$5.00), SWE-bench Verified 74.8%, OSWorld 2.0 desktop agent automation, and sub-80ms Ultrafast Mode.
| Metric | DeepSeek-R1 | GPT-6.1 Sol |
|---|---|---|
| SWE-bench Verified | 49.2% | 74.8% |
| MMLU-Pro | 84% | 92.4% |
| MATH-500 | 97.3% | 96.2% |
| Context Window | 128,000 tokens | 1,000,000 tokens |
| Input Pricing (per 1M) | $0.55 | $1.25 |
| Output Pricing (per 1M) | $2.19 | $5.00 |