google/gemma-4-26B-A4B-it logo

MODELS

google/gemma-4-26B-A4B-it

by Google

modellead-sourcehugging-face-popular-modelsgooglegemmagemma-4mixture-of-expertsmultimodalimage-text-to-textlong-contextapache-2.0hugging-facesource:blog.googlegemma-4-26b-a4b-itai-studiogemini-api
Visit WebsiteDocumentation

Overview

Google’s Gemma 4 26B A4B IT is a 26B-class Mixture-of-Experts Gemma 4 model variant for image-text-to-text workflows, with 25.2B total parameters, 3.8B active parameters, and a 256K context window.

Details

Gemma 4 26B A4B IT is part of Google’s Gemma 4 family. Google’s Gemma 4 model overview describes a 26B Mixture-of-Experts model designed for high-throughput advanced reasoning, with downloads available through Kaggle and Hugging Face. The official model card lists the 26B A4B MoE configuration with 25.2B total parameters, 3.8B active parameters, a 256K context window, text/image modalities, benchmarks, usage guidance, and Apache 2.0 licensing. The Hugging Face page for google/gemma-4-26B-A4B-it categorizes the artifact as image-text-to-text, and Google Gemini API release notes state that gemma-4-26b-a4b-it was released as part of Gemma 4 and is available in AI Studio and through the Gemini API.

When to Use

Use for evaluating a Google Gemma 4 Mixture-of-Experts model with text and image modalities. Use when you need a long-context Gemma model variant with a 256K context window. Use when comparing a 26B-class MoE model that activates 3.8B parameters per inference path. Use when you want an Apache 2.0-licensed Gemma 4 model available through official Google resources and Hugging Face.

Getting Started

  1. Read the Gemma 4 overview to understand the model family and official access paths: https://ai.google.dev/gemma/docs/core
  2. Review the Gemma 4 model card for parameters
  3. modalities
  4. benchmarks
  5. license
  6. and usage guidance: https://ai.google.dev/gemma/docs/core/model_card_4
  7. Inspect the specific model artifact on Hugging Face: https://huggingface.co/google/gemma-4-26B-A4B-it
  8. Check Gemini API release notes and pricing if you plan to use the hosted API path.
  9. Run evaluation prompts using representative text and image inputs before production use.

Key Features

  • •26B A4B Mixture-of-Experts Gemma 4 variant
  • •25.2B total parameters
  • •3.8B active parameters
  • •256K context window
  • •Text/image modalities
  • •Hugging Face category: image-text-to-text
  • •Available via AI Studio and Gemini API according to Google release notes
  • •Apache 2.0 license listed in Google’s model card

Capabilities

  • •image-text-to-text
  • •multimodal text/image processing
  • •long-context prompting
  • •Mixture-of-Experts inference
  • •advanced reasoning evaluation
  • •structured JSON output and function-calling workflows as described in Google’s Gemma 4 announcement

Last updated Jun 4, 2026