CISCai

/

Mistral-7B-Instruct-v0.3-SOTA-GGUF

like 1

The SOTA is about the GGUFs themselves, not about the model.

I'm using the term to denote that I'm only producing State-of-the-Art quantizations, ie only 4-bit and below with custom imatrices.

I also do a fair bit of testing, tuning and fixing, not just quant-and-dumps, all my GGUFs should be fully working, and even the 1-bit ones at least somewhat usable.