Compare

X vs Y

101 pairs. The pool is the 50 models offered by the most providers, then at most 8 models inside each family. Pairs stay inside that family. The URL puts the two ids in alphabetical order, and that URL is canonical.

Provider count is a coverage signal from Models.dev, not a quality score and not an OpenRouter ranking. The reference workload is 100 requests/day, 2,000 input tokens, 500 output tokens, no cache.

Claude

DeepSeek

GLM

Google

GPT

Kimi

Open weights

Qwen