Claude Opus 5.5 vs GPT-6.1 Sol: which one to call, and how to measure it on your own prompts