---
title: Julia-1 Decision Demo
emoji: 🧭
colorFrom: blue
colorTo: red
sdk: gradio
sdk_version: 6.28.0
app_file: app.py
short_description: "Typed decision model: state in, choices + probabilities"
python_version: "3.12"
startup_duration_timeout: 30m
license: apache-2.0
models:
  - SupersonicLabs/Julia-1
---

# Julia-1 · typed decision model

Demo of [SupersonicLabs/Julia-1](https://huggingface.co/SupersonicLabs/Julia-1), a
144M-parameter typed decision model: give it a *state* plus any set of named
questions — a **choice** between 2–20 candidate options, a rubric **score**, or a
yes/no (**noul**) — and it answers each one with a selection plus full
per-option probabilities. Multiple questions are scored independently in a
single pass.

The `julia/` package in this Space is vendored verbatim (Apache-2.0) from the
model repository; weights are pulled from the Hub at startup
(`snapshot_download`) and loaded with `julia.load_model(..., device="cuda",
strict_encoding=True, max_length=8192, head_length=512)`.

- **Input**: state/context text + a JSON object of named questions
- **Output**: per-question selected option, rubric score, or true/false —
  with probabilities, rendered as bars plus the raw JSON
- **Hardware**: ZeroGPU (`zero-a10g`); FP32 weights are 577 MB, inference runs
  in well under a second warm