llama.cpp ships /v1/systemone endpoint for local Jev-style decision models

mitsuhiko · x · 2026-10-03

ggerganov announced decision models are now available in llama.cpp: a new /v1/systemone endpoint in the latest builds enables Jev-style inference running locally, efficiently, and privately, with multiple open models supported and more coming. mitsuhiko responded positively but asked for model-type labeling and to rename the feature "classifiers" instead of "decision models."

Related event: llama.cpp adds decision models via /v1/systemone, returning option probabilities in a single forward pass(8 posts)→

Original post →

More from coding & agent

coding & agent channel →