Changing model in the previously sent message dropdown after a failed prompt does not apply the newly selected model
Steps to Reproduce
ask grok 4.7 anything
cursor claims its not in the registry (unrelated issue)
change the model to one that we know works
sending a message fails for the same issue around the old model
Expected Behavior
I’m fine with 4.7 not working because that model is not great but the dropdown should change the model. this is also relevant when there are outages on certain models.
the dropdown in the chat all the way at the bottom does work, if i change it there and press submit in the chat message its fine.
highlighting is not visible in whatever color scheme i have it in for some reason, cant tell when you are highlighting text in a previously sent message
Hi @joeybab3, the model picker on an edited message isn’t being applied in the IDE layout right now. The resend still goes out with the model set on the chat itself, which is why it kept failing on Grok 4.7. I’ve passed this to the team.
In the meantime, change the model in the dropdown under the chat input at the bottom (as you found), then resend or edit the message. That dropdown updates the chat’s model, so the resend picks it up.
The Grok 4.7 error looks like it may be tied to the 500K context option being selected while Max Mode is off. Opening Grok 4.7’s options and setting Context to 256K (or turning on Max Mode) should get it running again.
On the highlighting, could you raise a seperate thread with more information?
Yes to all, worth noting I never set the context to 500k tokens, that seems the be the default when trying to switch models? regardless an error message stating that it is due to the context length would probably be a better experience then the error message. Why is 500k even selectable if it requires another setting to be enabled? I would say if this is something that requires max mode have it be shown but greyed out, and certainly should not be able to become the default.
Hi @joeybab3, you shouldn’t need to do anything more. Setting Context to 256K for Grok 4.7 in the picker under the chat input replaces the 500K that had been saved for it, so since that worked for you the old value is gone.
If it ever comes back:
Open the model picker under the chat input, hover Grok 4.7 and click the reset arrow (or Restore defaults in its options, depending on your version). That clears its saved settings and goes back to the default, 256K.
Older chats remember their own model settings, so if one of them still fails, pick the Context again in that chat before sending.
On your plan, 500K doesn’t need a separate Max Mode setting, so picking it fresh should work. The failures came from an older 500K choice being sent without the setting that goes with it.
