I have a small MCP server I use to deal with administrative tasks. I have found several LLMs work to use the MCP server—so I can choose between GGUF files on Hugging Face and use whichever one works best (or fastest).
Today support for Ling models (made by Ant Group, somehow related to Alibaba) was added to llama-cpp
. So I downloaded Ling-3.0-tiny-Q4_K_M.gguf
and tried it out on my MCP server.
• It works great—very fast, accurate tool calling.
• It can likely completely replace other LLMs such as LFM 2.5 that I usually use.
• It seems to me that LLMs are quickly becoming commodities—many of them can be used for a certain task, and they can be replaced easily.
I am not sure if I will use Ling 3.0 Tiny for daily tasks, or if I will stick to LFM 2.5, but having the choice has value to me. I won't be stuck on one provider or paying monthly fees to use a specific model.