sdk.vercel.ai is also a nice playground where multiple models can be used for interaction at the same time. Regarding this application scenario, is it used to compare the quality of model responses? Although there are many large model evaluation organizations now, for ordinary users, common standards often fail to guide the correct choice, after all, everyone's use case is different.

OpenRouter is really good for testing out different models. It's pay as you go for models that are usually behind $20/month subscriptions

Obligatory Ollama link: https://ollama.com/

May be of interest.

Check out Ollama's Modelfile, too: https://github.com/ollama/ollama/blob/main/docs/modelfile.md