About this chart
Best for the costOn the line, but beatenBeaten on score and priceOn OpenRouter Intelligence, Agentic and Vision averaged, each rescaled across only the models all three rank, at each model’s best setting. 0 is the ranking’s 2nd percentile, 100 its strongest; the table keeps the raw score. 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