Report
Decision models arrive in llama.cpp, with a Julia-1 playground to try them
The llama.cpp announcement says `/v1/systemone` supports multiple open models; the playground’s creator says Julia-1 ranks options without generating text.
TLDR
An October 2 announcement says the latest llama.cpp builds offer a /v1/systemone endpoint for local, efficient and private Jev-style inference, supporting multiple open models with more planned. In a separate October 5 post, a user says they built a Gradio playground where Julia-1 ranks options, estimates a yes/no probability and scores an ordered scale without generating text. Changing the context lets visitors see how its decisions shift.
Combined views
4.8K
2 Sources, first seen ago
