Skip to content

Llama-3-Patronus-Lynx-8B-Instruct

An open model that checks answers for RAG hallucinations

Visit official site

What Llama-3-Patronus-Lynx-8B-Instruct is

The model is described as a fine-tune for hallucination detection in RAG setups that judges faithfulness to a document.

What you can do with it

  • Prepare question, document and answer text
  • Fill them into the recommended prompt format
  • Call the pipeline and read the JSON result

Pros & cons

✓ What we like

  • Noted for FinanceBench and CovidQA results
  • cc-by-nc-4.0 licence allows non-commercial use

! What to watch out for

  • The licence is limited to non-commercial use
  • Evaluation needs a question, a document and an answer

FAQ

What does the model output?

A JSON result with REASONING and SCORE.

How is it called?

Through the Hugging Face pipeline with a recommended prompt.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review