openjev/openjev
Zero-Shot Classification • 27B • Updated • 9.86k • 177
27B decision model. Text or a screenshot plus your options in, a scored decision out, in one pass.
Note BF16 weights for text and screenshots.
Note GPU checkpoint for text and screenshots.
Note llama.cpp builds. Q4_K_M runs on a 24 GB card.
Note 8-bit text decisions on Apple silicon.
Note Text decisions on Apple silicon in about 15 GB.
A decision API over any open model, one pass per question
Note Serve the same decision API on your own GPU or Mac.