DiffusionGemma decision endpoint
Free, rate-limited Hugging Face endpoint serving DiffusionGemma decisions in the /v1/systemone format without fine-tuning.
Overview
This Hugging Face Space is a guide page for a free, public, rate-limited DiffusionGemma 26B-A4B endpoint. It serves structured reads in a Jev-style decision API alongside OpenAI-compatible text generation on Hugging Face Inference Endpoints, using a vLLM pull request that adds the feature.
According to the list summary, it reads a probability for every answer from DiffusionGemma without fine-tuning, behind the /v1/systemone wire format. It is useful for trying Jev-compatible client code against an open model at no cost, within rate limits. Because the endpoint shares the /v1/systemone request shape, the same state-and-questions payloads used with hosted Jev can be sent to it unchanged, which makes it a convenient way to compare an untuned open model with Jev on your own questions. The Space itself is a static guide page under an MIT licence.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Any typed question sent in /v1/systemone format(inferred) | — |
How it works
- Point a Jev-compatible client at the endpoint
- Send state and typed questions
- Receive per-option probabilities
Strengths
- Free to try
- Jev-compatible wire format
- Also serves text generation
Caveats
- Rate-limited public endpoint
- Based on an unmerged vLLM change at the time
- Minimal documentation
Getting started
See the Space at huggingface.co/spaces/victor/DiffusionGemma-free-endpoint.At a glance
- Type
- Service
- Status
- prototype
- Runtime
- hosted endpoint (vLLM)
- License
- MIT
Use cases and integrations
How curated lists describe it
Free Hugging Face Space that reads a probability for every answer from DiffusionGemma, without fine-tuning, behind the `/v1/systemone` wire format
AnotiaWang/awesome-decision-models
Frequently asked questions
Is it free?
Yes, but rate-limited.
Which model?
DiffusionGemma 26B-A4B, without fine-tuning.
Can I use the TypeSafe SDK?
It uses the /v1/systemone wire format, so Jev-compatible clients should work.
Related projects
- TypeAR★ 946
TypeLLM: Jev-inspired type-safe decoding for existing autoregressive LLMs on SGLang, with schema-guaranteed typed outputs.
ProjectbetaPython llama.cpp-based C++ server that turns any LLM into a System One decision model, reading answers from next-token logits.
ProjectbetaC++MLX library that turns an existing local model into a decision engine by reading candidate logits, with no fine-tuning, on Apple Silicon.
ProjectprototypePythonPython library that turns an open LLM into a Jev-style classifier via constrained single-token choice indexes in one forward pass.
ProjectprototypePythonOpen Jev-style decision model built on Gemma 3 270M with a scoring head, scoring all options in one forward pass on CPU.
ProjectresearchPythonOfficial code for the FastJEV paper: training-free reduction of redundant candidate computation in multimodal JEV-style models.
ProjectresearchPython