featherless-ai/gemma-4-26B-A4B-classifier

Hugging Face
VISIONPricing:Input $0.28 / Output $0Concurrent Unit Cost:2Model Size:26BQuant:FP8Context Size:32k Classification model Architecture:Transformer Warm

The featherless-ai/gemma-4-26B-A4B-classifier is a 26 billion parameter Gemma-based model developed by Featherless AI, specifically designed for structured classification tasks. It leverages the Simple Jev classifier API to convert text, conversation, or image context into structured decisions like choice selection, rubric scoring, or proposition judgment. This model excels at applications requiring precise, programmatic outputs such as intent detection, content moderation, and support routing, supporting a 32768 token context length.

Loading preview...

Model Overview

The featherless-ai/gemma-4-26B-A4B-classifier is a 26 billion parameter Gemma-based model developed by Featherless AI, optimized for structured classification tasks. It utilizes the Simple Jev classifier API to transform various inputs into programmatic JSON outputs, making it suitable for direct application integration. This model supports a 32768 token context length and is currently available in beta on Featherless developer plans.

Key Capabilities

  • Structured Decision Making: Converts text, conversation history, or image context into structured decisions.
  • Flexible Question Types: Supports three primary question types:
    • choice: Selects from named candidates with confidence and probabilities.
    • score: Provides an expected rubric index, confidence, and probabilities for ordered rubrics.
    • noul: Judges a yes/no proposition with a confidence score.
  • Multi-modal Context: Capable of processing text, structured data, chat history, and images/vision inputs.
  • API-driven: Designed for integration via a dedicated /v1/classifier endpoint, returning JSON outputs.

Good For

  • Intent Detection: Automatically categorizing user queries or messages.
  • Content Moderation: Classifying content based on predefined criteria.
  • Support Routing: Directing customer inquiries to appropriate departments.
  • Relevance Scoring: Evaluating the relevance of information or documents.
  • Visual Classification: Analyzing images to make structured decisions (e.g., identifying objects, assessing scenes).
  • Automating Agent Actions: Determining the next best action for an AI agent based on context.

This model provides a robust solution for applications requiring precise, programmatic classification, offering a clear advantage over general-purpose LLMs for specific decision-making workflows. For more technical details, refer to the Simple Jev GitHub repository.