Claude 3.7 Sonnet vs GPT-6.1 Sol
A direct, empirical evaluation of Claude 3.7 Sonnet (Anthropic) against GPT-6.1 Sol (OpenAI).
Claude 3.7 Sonnet
Developed by Anthropic · Proprietary parameters
Enterprise software engineering, full-repo refactoring, and agentic workflows.
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 | Claude 3.7 Sonnet | GPT-6.1 Sol |
|---|---|---|
| SWE-bench Verified | 70.3% | 74.8% |
| MMLU-Pro | 82.1% | 92.4% |
| MATH-500 | 96.2% | 96.2% |
| Context Window | 200,000 tokens | 1,000,000 tokens |
| Input Pricing (per 1M) | $3.00 | $1.25 |
| Output Pricing (per 1M) | $15.00 | $5.00 |