Not sure exactly what the solution is but say I am using gemini and it returns a poor answer I would like a way to easily try a different model.
Harlequin Orbitron
Bump
Lavender Chipmunk
+1 on this
the main use case i see is that sometimes i will switch to a premium model or a slower reasoning model for a more complex query. however, when i open a new conversation, most of the time, i want to go back to a quicker and non-premium model to save the premium credits and get quicker response times.
Teal Dew
Definitely this, I have wasted premium queries on “stupid“ questions already. If I want a premium model, I want to *ask for it*, not have it be default.
Purple Star
oh, were you thinking about this on a “per conversation” or a “per message” basis?
figuring out if my post is a duplicate of yours, didn’t find yours during my first search:
To add on this, I'd like to have a profile which focuses much more on UX research/reasoning and another for more day to day tasks. Having profiles/spaces would be nice to achieve this (if i understood this feature request correctly)
Copper Pickle
Also noticed that even disabling gemini as an eligible model it keeps getting selected as the default. This change would be awesome.
Blue Puzzle
Any updates on this? I’d like to have a bunch of Custom Instruction Presets (?) that I can pick. Could work similar to the model picker, as a little menu inside of the input box.
Rose Stapler
imo It would be really cool to break down the option as follows:
- Choice of default model for new conversations
- Choice of default model for requests via URL (use with unduck, for example)
Each of these options would have a default “use last model used” option to simplify things.
it would be really cool to see this feature added, because I'd like to have Gemini 2.5 Thinking with Grounding when I do a query with Unduck, but I like to use 2.5 Pro when I go to a normal conversation
Mark Florkowski
Team
Choice of default model for requests via URL - each model has a designated url:
Rose Stapler
oh neat, I hadn't seen that, I'll add it to my search engines right away lol.
however, if I may defend my opinion, I see a few downsides if I compare this to a kind of dedicated selector:
- firstly, I can't see how to activate search grounding. Being able to do a kind of internet search in this way would be a lot like perplexity, but without the bloat and with the ux of T3 chat (this one is not necessarily related to the selector feature itself, but yeah, it'd be cool to have a feature like this in the search bar).
- also, this function doesn't really integrate with unduck. Lastly, I've found myself using it a lot to prompt different things in addition to my traditional searches, but going from a !yt to a @t3 (if i want to have my default model selected on firefox) isn't really consistent (it's not a big deal, but on an azerty keyboard it's not as
convenient).
it might seem a bit specific, but a function implemented like that would make things a bit more seamless for some queries
Mark Florkowski
Team
firstly, I can't see how to activate search grounding.
add &search=true to enable search for the request. Note that it will not remain enabled, but you can always toggle it back on when you continue. Most of the time, it’s not required to search on every message
> this function doesn't really integrate with unduck. Fair, I believe there is a community for of unduck with more customizations. I cannot remember which one off the top of my head though
Rose Stapler
Oh that's really cool to know.
I seem to have missed a few chapters here lol.
Thanks for your replies!
Mark Florkowski
Team
No problem! We should probably make this information more discoverable, i just have not had a chance to figure out where to put it!
Purple Kangaroo
To add to this, when using bangs from unduck it would be nice have a query param `model=search-default` that would pick the default model you want instead of the last model used. I’ve wasted multiple reasoning model questions because of this. Would be nice to also maybe have a profile or config of some kind so `config=search-config` would pick a light weight model with search enabled or something
Coffee Moss
Choosing a default model for new conversations would stop me from trying to ask a question to nano banana pro…
Log in to comment and vote
Comments17
Purple Kangaroo
Not sure exactly what the solution is but say I am using gemini and it returns a poor answer I would like a way to easily try a different model.
Harlequin Orbitron
Bump
Lavender Chipmunk
+1 on this
the main use case i see is that sometimes i will switch to a premium model or a slower reasoning model for a more complex query. however, when i open a new conversation, most of the time, i want to go back to a quicker and non-premium model to save the premium credits and get quicker response times.
Teal Dew
Definitely this, I have wasted premium queries on “stupid“ questions already. If I want a premium model, I want to *ask for it*, not have it be default.
Purple Star
oh, were you thinking about this on a “per conversation” or a “per message” basis?
figuring out if my post is a duplicate of yours, didn’t find yours during my first search:
https://feedback.t3.chat/p/presets-concise-responses-personas-tone-and-style-aka-help-me
Harlequin Bunny
I was mostly thinking per conversation but since t3 allows to switch models mid conversation, setting profile per message won’t be bad either
useful when context are closer but still changing in the conversation
like going from frontend → backend → devops
this can go to the next level by adding automatic profile selection based on the message content
something similar to
https://feedback.t3.chat/p/choose-the-model-dynamically
Yellow Nucleon
To add on this, I'd like to have a profile which focuses much more on UX research/reasoning and another for more day to day tasks. Having profiles/spaces would be nice to achieve this (if i understood this feature request correctly)
Copper Pickle
Also noticed that even disabling gemini as an eligible model it keeps getting selected as the default. This change would be awesome.
Blue Puzzle
Any updates on this? I’d like to have a bunch of Custom Instruction Presets (?) that I can pick. Could work similar to the model picker, as a little menu inside of the input box.
Rose Stapler
imo It would be really cool to break down the option as follows:
- Choice of default model for new conversations
- Choice of default model for requests via URL (use with unduck, for example)
Each of these options would have a default “use last model used” option to simplify things.
it would be really cool to see this feature added, because I'd like to have Gemini 2.5 Thinking with Grounding when I do a query with Unduck, but I like to use 2.5 Pro when I go to a normal conversation
Mark Florkowski
Choice of default model for requests via URL - each model has a designated url:
Rose Stapler
oh neat, I hadn't seen that, I'll add it to my search engines right away lol.
however, if I may defend my opinion, I see a few downsides if I compare this to a kind of dedicated selector:
- firstly, I can't see how to activate search grounding. Being able to do a kind of internet search in this way would be a lot like perplexity, but without the bloat and with the ux of T3 chat (this one is not necessarily related to the selector feature itself, but yeah, it'd be cool to have a feature like this in the search bar).
- also, this function doesn't really integrate with unduck. Lastly, I've found myself using it a lot to prompt different things in addition to my traditional searches, but going from a !yt to a @t3 (if i want to have my default model selected on firefox) isn't really consistent (it's not a big deal, but on an azerty keyboard it's not as
convenient).
it might seem a bit specific, but a function implemented like that would make things a bit more seamless for some queries
Mark Florkowski
add
&search=trueto enable search for the request. Note that it will not remain enabled, but you can always toggle it back on when you continue. Most of the time, it’s not required to search on every message> this function doesn't really integrate with unduck.
Fair, I believe there is a community for of unduck with more customizations. I cannot remember which one off the top of my head though
Rose Stapler
Oh that's really cool to know.
I seem to have missed a few chapters here lol.
Thanks for your replies!
Mark Florkowski
No problem! We should probably make this information more discoverable, i just have not had a chance to figure out where to put it!
Purple Kangaroo
To add to this, when using bangs from unduck it would be nice have a query param `model=search-default` that would pick the default model you want instead of the last model used. I’ve wasted multiple reasoning model questions because of this. Would be nice to also maybe have a profile or config of some kind so `config=search-config` would pick a light weight model with search enabled or something
Coffee Moss
Choosing a default model for new conversations would stop me from trying to ask a question to nano banana pro…