Incorrectly finding /think tag for GLM 4.5 models leading to empty message
When I ran a quick fork prompt on GLM 4.5 from a different thread (to compare answers between LLMs), it appears as though t3.chat didn’t properly find the </think> tag, and included the actual response in the ‘reasoning’ section. don’t think it’s a critical bug, but it does disable things like copying the results out (which is pretty annoying in this case). Really bad timing as I am trying to show off T3.chat to a bunch of Civil/Environmental engineers here!
Thread ID is listed below.
Log in to comment and vote
Comments7
Yash Singh
Sorry about that! Should be fixed now.
Amber Fuelcell
still happening
Amber Fuelcell
the issue is that the response is rendered and once it pull the grounding sources it goes back to an empty message
Yash Singh
Could you send link the thread with the issue so that I can take a look?
Amber Fuelcell
https://t3.chat/chat/111a1ddb-8d89-4adc-a9fb-e5009ad48fbb
Amber Fuelcell
I am sorry but this is still happening is there a timeline on when this will be fixed, I mainly use GLM as it gives better results due to it always searching. PLEASE GET THIS FIXED ASAP
Amber Fuelcell
claude fails when it wants to do 2 searches :[