Claude Opus vs GPT-6.1 Sol
In Nivana’s weekly evals Claude Opus (Anthropic) beats GPT-6.1 Sol (OpenAI) on 7 of 8 task types and loses on 1. Claude Opus costs $4 in / $20 out per 1M tokens; GPT-6.1 Sol costs $2 in / $10 out per 1M tokens.
| Task | Claude Opus | GPT-6.1 Sol |
|---|---|---|
| quick answers and everyday questions | 83 | 84 |
| writing | 95 | 89 |
| coding | 97 | 88 |
| reasoning and analysis | 98 | 82 |
| maths | 93 | 77 |
| structured data and JSON | 88 | 86 |
| long documents | 91 | 86 |
| translation | 91 | 83 |
With Nivana you don’t choose between them: the router sends each request to whichever wins for it.