Agreed, would love this feature. Cursor has this and it’s been great not having to think about which model is used.
The question is, would we want this to run on every user input and auto-detect the best model based on the previous context/answers or just select one based on the first questions asked by the user just like how the auto-titles work on most ai chats.
Potential downside to this approach is that it will be more expensive and slower.
Harlequin Bunny
how about \@ model or ddg bang to select model as you prompt as temporary solution
Mark Florkowski
Team
I would like to have ddg-style bangs regardless
Blue Storm
Could possibly using a like/dislike rating system based on the responses given to train
Amber Beach
Thats fire, training what kind of LLM responses you like the most
Teal Savanna
This is cool, or just having a small text next to the recommended model so we know what should we use even with no idea
Salmon Jaguar
Something like what notdiamond have is also cool
Amber Beach
Yeah this look good!
Chocolate Aurora
Didn’t care about this until the speed of model releases got ridiculous!
I’d like to prioritize my searches like:
base query: Latest Gemini Flash Model
thinking model: Latest Gemini Pro model
search model: Latest Gemini Flash Model
Or something like that. Or just pick whatever you think is best! I don’t want to pick anymore.
Log in to comment and vote
Comments12
Silver Pretzel
Agreed, would love this feature. Cursor has this and it’s been great not having to think about which model is used.
The question is, would we want this to run on every user input and auto-detect the best model based on the previous context/answers or just select one based on the first questions asked by the user just like how the auto-titles work on most ai chats.
Potential downside to this approach is that it will be more expensive and slower.
Harlequin Bunny
how about
\@ model or ddg bang to select model as you prompt
as temporary solution
Mark Florkowski
I would like to have ddg-style bangs regardless
Blue Storm
Could possibly using a like/dislike rating system based on the responses given to train
Amber Beach
Thats fire, training what kind of LLM responses you like the most
Teal Savanna
This is cool, or just having a small text next to the recommended model so we know what should we use even with no idea
Salmon Jaguar
Something like what notdiamond have is also cool
Amber Beach
Yeah this look good!
Chocolate Aurora
Didn’t care about this until the speed of model releases got ridiculous!
I’d like to prioritize my searches like:
base query: Latest Gemini Flash Model
thinking model: Latest Gemini Pro model
search model: Latest Gemini Flash Model
Or something like that. Or just pick whatever you think is best! I don’t want to pick anymore.
Purple Kiwi
This could be a model addition - https://openrouter.ai/docs/features/model-routing#auto-router. Not Diamond handles the model selection under the hood. They support thumbs up / thumbs down for training as well, but that can come later.
Harlequin Bunny
I don’t really care what specific model is being used. Auto would make my life easier
Beige Mirror