About this chart
Best for the costOn the line, but beatenBeaten on score and priceOn OpenRouter Intelligence and Agentic averaged, each as a percentile among only the models both rank, at each model’s best setting. USD per million tokens, three input to one output. Higher reasoning settings spend more tokens, so cost more per call.
Intelligence and speed from Artificial Analysis, other rankings from Arena (CC BY 4.0). Prices from OpenRouter. Refreshed every 24 hours. /api/frontier