"Fastness" scale

Issue
We can currently already filter by ‘Fast’ in the model picker… but unironically, all models are tagged as ‘fast’. Would be cool to have some sort of scale of how fast a model is.

Why is this important?

Often, I ask very simple questions that any LLM can answer me. I don’t need Claude 3.7. I just need it to be quicker than Google. And such a scale would help me weigh how long I’m willing to wait for an answer vs how complex the query is.

More importantly, it would be good to see how long a model takes to spin up before generating an answer. Gemini 2.5 Pro regularly takes ~10 seconds before I see the first token, while Fireworks v3 is near-instant. This makes a huge difference in how… likely I am to use a model…

Implementation Suggestion (might suck)
Would be cool to have some sort of scale (maybe a small ‘5s‘, ‘3s’, … chip that denotes how long it takes to generate an answer to a reference question?) next to it that denotes how quick the model is?