You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Llama cpp supports reasoning budgets that only apply to the reasoning part and can cut off reasoning with a predefined string. That would be awesome to have in LM Studio.
Reacted by ian-morgan99, izMoff, banan, goodglitch, FreeeFry, Sh1d0w, homerunner55 and William Soto
--reasoning-budget N
token budget for thinking: -1 for unrestricted, 0 for immediate end, N>0 for token budget (default: -1)
(env: LLAMA_ARG_THINK_BUDGET)
--reasoning-budget-message MESSAGE
message injected before the end-of-thinking tag when reasoning budget is exhausted (default: none)
(env: LLAMA_ARG_THINK_BUDGET_MESSAGE)
That works for now, but I assume it is globally applied? I use different models for different workflows. So it means I have to change the variables and restart lm studio every time I change workflow, not ideal, but still very worthwhile for me. Thank you!
There should really just be a command line field in model settings, where we could input any additional arguments that llama.cpp supports, instead of only having access to whichever ones LM Studio devs deemed important enough to be added to UI.
Reacted by takeseem, David Schor, Christopher J. Steele, Brano Pavelka, lockotar, Ankit Soni, FreeeFry, goodglitch and Malte Koitka
Activity
DavidSchor commentedon May 28, 2026
I am referring to this:
--reasoning-budget N
token budget for thinking: -1 for unrestricted, 0 for immediate end, N>0 for token budget (default: -1)
(env: LLAMA_ARG_THINK_BUDGET)
--reasoning-budget-message MESSAGE
message injected before the end-of-thinking tag when reasoning budget is exhausted (default: none)
(env: LLAMA_ARG_THINK_BUDGET_MESSAGE)
https://github.com/ggml-org/llama.cpp/blob/master/tools/server/README.md
takeseem commentedon May 28, 2026
There is a workaround:
✔️ [BETA] Enable LM Studio Engine Protocol
DavidSchor commentedon May 28, 2026
That works for now, but I assume it is globally applied? I use different models for different workflows. So it means I have to change the variables and restart lm studio every time I change workflow, not ideal, but still very worthwhile for me. Thank you!
izMoff commentedon May 30, 2026
There should really just be a command line field in model settings, where we could input any additional arguments that llama.cpp supports, instead of only having access to whichever ones LM Studio devs deemed important enough to be added to UI.