Which model should I use?

Quality is the model's rating on the Arena (LMArena), where people compare answers blind. For each task we use the Arena leaderboard that matches it (coding, math, creative writing, long prompts, instructions, images). Where a model has no rating for the task, its overall rating minus 25 points is used and the pick says so.

Price is what OpenRouter lists, blended three parts input to one part output, the usual mix for chat and coding. Models with no listed price are only offered if you can download them.

The score is a mix of the two that you steer with “What matters more?”. It is a guide, not a verdict: the Arena measures what people prefer in chat, not your data. Try your top three on your own prompts before you commit.