Reasoning not working
Reasoning is not working. No reasoning is happening, but otherwise the model works fine.
I am using the new llama.cpp Web UI.
Is there something wrong with the chat template?
i run it with:
llama-server -hf unsloth/Ministral-3-14B-Reasoning-2512-GGUF:Q8_0
--jinja --temp 0.7 --top-k -1 --top-p 0.95 -ngl 99
--ctx-size 64000
--cache-type-k q8_0
--cache-type-v q8_0
--flash_attn on
that works, but the reasoning is questionable. Sometimes excessive, in followup questions it fails to use the thinking tags (or llama.cpp webui fails to format em right). Seems unstable
Also experiencing this exact thing.
It seems to be working for me, can you guys use the latest llama.cpp update and see if it's improved? apparently mistral 3 models were resolved.
See these 2 pull requests and issues:
https://github.com/ggml-org/llama.cpp/pull/17945
https://github.com/ggml-org/llama.cpp/issues/17980
I did a git pull today 15 december 2025 and was able to load the model using llama-server. Tested it on reasoning tasks and works pretty well.