Model Selection Guide

Noticia Go provides access to two AI language models. Each model has different characteristics that make it more suitable for certain types of legal document review tasks. Both models operate on structured text inputs and can assist with relevance review, privilege classification, objective coding, personally identifiable information (PII) detection, and other text analysis tasks.

Available Models

Meta LLaMA 3.3 Instruct (70B) [Default]

• Best suited for: Tasks that require structured classification, such as binary tagging (e.g., privileged vs. non-privileged documents) and clearly defined labeling.

• Strengths: Performs reliably with detailed prompts; maintains consistency across repeated classification tasks; handles large context windows well compared to smaller models.

• Limitations: May struggle with nuanced, ambiguous, or highly context-dependent text; less effective in tasks requiring inference beyond explicit statements.

OpenAI OSS 120B

• Best suited for: High-throughput processing, lightweight extraction, or bulk tagging where speed is a priority.

• Strengths: Fast and efficient for simpler, well-defined tasks; can handle large volumes of text.

• Limitations: Less capable of interpreting complex context or subtle language; may miss edge cases or nuanced legal distinctions in documents.