4 ms·

I inspected manually and indeed the 27B is doing worse, but I believe it could be due to the exact GGUF in the ollama repository and/or with the need of adjusti

by antirez 7mo ago

I inspected manually and indeed the 27B is doing worse, but I believe it could be due to the exact GGUF in the ollama repository and/or with the need of adjusting the parameters. I'll try more stuff.

andhuman 7mo ago
Isn’t llama.cpp’s implementation of Qwen 3.5 better, or am I misinformed?
- antirez 7mo ago
  There was a recent fix by ollama and I used it.