Skip to main content

Better Model Picker

  • Improved search/filtering

  • Organize by categories

  • Grid view is not very useful

  • Better pinned/favorites system

Status: Completed13 comments

Log in to comment and vote

Comments13

  • Lime Sunbeam

    • Lime Sunbeam

      Specifically, in the Settings > Models
      All models should be enabled by default NGL..

      Seriously I always immediately enable Every new model because I’m constantly comparing outputs across multiple models for one reason or another.

      Instead of enabling models, we should just be given a normal Favorites tab on top of the normal model picker.

      Curtesy of Nanobanana pro, take your pick man

      I’m partial to the top right and bottom left concepts

      Either the bottom left, or merge the top 2 concepts actually and it would be great.

      though the bottom left is simply the superior arrangement if you make the provider option a drop down / expanding field with box checks instead of just choosing 1 option.


      In Settings > Models. what should actually be there is a favorites check box instead of toggle on/off. or just adding it in. Since I know other people would want the option to disable models.


      Honestly, I really feel like the picker on the main page is just too minimal. Minimalism is great, but only if you know what should and shouldn’t be present. And this is simply far, far, far too minimal.

      The current picker being just a single scrollable list leads to many many many fat finger accidents where I accidentally click Claude 4.5 hiku without reasoning, when I meant to click the Reasoning model….

      While sure, that’s a skill issue, And definitely because I always turn on every model thus bloating my list. The real reason for why it happens is not because I’m clicking an option in a list. The reason I miss click is because I’m going from this long tedious scroll and then shifting gears to go click after frustratingly trying to search by eye for the single model in my list.

      And again, I just simply turn on all models regardless because I’m always shifting between them for different reasons. Like comparing, or requiring different features in the same chat string, so I need to go from claude to get a description to say nanobanana to turn said description into a picture.

      I don’t wanna jump between 5 different chats copy pasting things.

  • Gold Orbit

    Show max context window each model in selector

  • Jade Lander

    Good to ba able to create own group like , work , dotnet , School and life

  • Black Sorbet

    •

    Merged request

    Separate section for flagship models

    I find myself having to unpin and repin whenever a lab drops a better model. I’d like to be able to easily choose from the flagship models from each provider. Or have some way to easily identify the “best” / most efficient ones maybe a label for each of these categories from each provider. best premium and non-premium

    • Mark Florkowski

      Team

      Unfortunately newest offering does not necessarily mean “best model”. It is highly dependent on your particular work, which is why we let users pin the models that they want to pin.

      • Black Quasar

        makes sense, I find myself having to search for which models are comparable like in artificial analysis and which are the fastest. Maybe some of these scores could be displayed or linked to or something?

        • Mark Florkowski

          Team

          Yup, definitely something I want to do!

  • Copper Honey

    •

    Merged request

    Degraded Badge Positioning in List is off

    Images tell more than words:

  • Salmon Pickletop

    The pinned system needs improvement for finding the models you want in the Settings page, the UI is very spaced out for a list with >90 models in it.

    • Salmon Pickletop

      Or, remove the page entirely and direct you to use the regular model picker to pin models. It’s much nicer there. Just noticed you could do that.

  • Beige Rain

    It would also be nice to have some sort of continuously updated, curated list of newest models from various labs, ranked or compared against each other. It would be nice to avoid a ton of trial and error and just know what people think are the current top picks for various use cases.

    • Mark Florkowski

      Team

      The new Legacy designation should handle this