LocalLLaMA
QATs Q4_0 from Google have more precision than Q4_K_XL from Unsloth (at least some)
I wanted to try new QATs and opened two collections on HF (which HF found for me): https://huggingface.co/collections/google/gemma-4-qat-q4-0 https://huggingface.co/collections/unsloth/gemma-4-qat One strange thing caught my attention, for e.g. E4B: https://huggingface.co/google/gemma-4-E4B-it-qat-q4_0-gguf/resolve/main/gemma-4-E4B_q4_0-it.gguf 5.15 GB https://huggingface.co/unsloth/gemma-4-E4B-it-qat-GGUF/resolve/main/gemma-4-E4B-it-qat-UD-Q4_K_XL.gguf 4.22 GB How can _0 be la...
Read the full discussion on Reddit
This article was aggregated from Reddit. Click to join the conversation.
View on Reddit