Skip to main content

Model ranking analytics

Love the product. With it being so easy to switch models it would be realle nice if you could integrate at-a-glance model information for the most popular intelligence index categories directly into the model selection interface, a la https://artificialanalysis.ai/. This would provide users with an immediate, data-driven understanding of a model's performance characteristics across various benchmarks, aiding in more informed model selection.

Status: Completed11 comments

Log in to comment and vote

Comments11

  • Sapphire Crab

    I had a similar feature request that got merged into this one. I think the feature requested here would be kind of neat, but I personally would be satisfied with something much simpler. I’m not doing anything fancy and don’t need complex analysis of the relative merits of different models, I just want “for coding use X, for research use Y, for emails use Z”. Updated once a month or whenever new models are added.

  • Bronze Fox

    Community Rankings on different topics like coding, writing, research, reasoning, etc, would be really cool. Being able to filter models based on performance would make picking one of the 30 models so much easier. Until then, I’ll just Gemini 2.5 it up, lol

    • Aqua Moose

      Hard agree! Don’t know if the team has analytics set up, but a popularity rank would be useful. I would still prefer the model performance ranking if I had to choose though

  • Bronze Lunchbox

    Just came to suggest this. I’m a technical user, but there are so many models that the capabilities of each is sometimes too much to know by the top of my head.

  • White Sunshine

    As a new user I have no clue when to switch models, whould it be possible to have suggestions based on the input? (even if its async)

    • Mark Florkowski

      Team

      Yup! This is something we have planned, including an “auto” model that picks the best model for you automaticallt

  • Aqua Moose

    I feel this is essential. With new models coming left and right is hard to keep track if I should from the one I’m using. I think the simplest set of information should be a ranking of speed and “intelligence”. While reasoning strength is important some times you just need a good enough model with fast responses

  • Turquoise Deer

    •

    Merged request

    Leaderboard of popular models

    I’d like to be able to see which models are most popular. Within the last week, which models are the most used? (Ideally we’d also be able to rate responses, which would enable a leaderboard of models that have the best rated responses.)