[codelion]

There are ways to improve the performance of local LLMs with inference time techniques. You can try with optillm - https://github.com/codelion/optillm it is possible to match the performance of larger models on narrow tasks by doing more at inference.