Bonsai-JEV
Ternary-Bonsai-2-27B-gguf fine-tuned to return typed decisions, 27B parameters.
Community card
What this is
Bonsai-JEV is a typed-decision model: given a document and a closed set of options, it returns one of those options with a probability across the whole set rather than free text.
Its card lists training data: SargeDev/jev-distill-corpus-v3. That matters when you read any benchmark number about it, because a model tuned on a benchmark's own training split has an advantage over one that has never seen it.
How to run it
llama.cpp
GGUF runs on CPU. No Python, no GPU, and no network call in the path of a decision. Full llama.cpp guide.
llama-cli -hf dakebakery/Bonsai-JEVA public number is a shortlist, not a decision
The only benchmark that settles which model you ship is one built from your own historical decisions. We freeze that test set, run the candidates against it, and hand back the threshold your error budget supports.