v16 - Reasoning despite setting --reasoning off in llama.cpp
am i doing something wrong or is this expected behaviour? it is having a lot of problems with tool calling, tested web search and browser navigation in hermes so far, with qwen3.6 35b. and i see very often the thinking blocks, although i turned thinking off by setting --reasoning off in llama.cpp command.
Edit: the thinking blocks dont appear at all with the built in template. using the UD-IQ4_NL-XL quant
Edit: the thinking blocks dont appear at all with the built in template. using the UD-IQ4_NL-XL quant
Solved in the final v16 release which I have now promoted to official release.
great! thanks for the response. i tried it and it looks better when reasoning is on, but when reasoning is off, it fails tool calls and just stops working with API call errors