> I recently discovered that ollama no longer uses llama.cpp as a library, and instead they link to the low level library (ggml) which requires them to reinvent

> I recently discovered that ollama no longer uses llama.cpp as a library, and instead they link to the low level library (ggml) which requires them to reinvent a lot of wheel for absolutely no benefit (if there's some benefit I'm missing, please let me know). Here is some relevant drama on the subject: https://github.com/ollama/ollama/issues/11714#issuecomment-3172893576 https://github.com/ollama/ollama/issues/11714#issuecomment-3...