Skip to content

gemma 4 26B A4B it

google/gemma-4-26B-A4B-it
Open Source · chat · open-weights
GA Alert me on changes
Context
262.1K
Max output
32.8K
Input $/1M
$0.06
Output $/1M
$0.33
Modalities
text · image
Released
11 Mar 2026
License: apache-2.0 · google/gemma-4-26B-A4B-it
AI summary
● machine-written

Google Gemma 4 26B A4B: 26B MoE model with 262K context window

Gemma 4 26B A4B is an instruction-tuned Mixture-of-Experts model from Google with 25.2B total parameters, activating only 3.8B per token. It supports multimodal input (text, images, video) with a 262K token context window, native function calling, and structured output. The model is available under Apache 2.0 license across 11 providers.

What's new
  • 262K token context window with native function calling
  • Multimodal input support: text, images, and video (up to 60s at 1fps)
  • Configurable thinking/reasoning mode and structured output support
  • Mixture-of-Experts design: 3.8B active parameters per token from 25.2B total
Best for
Long-context document analysis and summarizationMultimodal tasks with images and videoCost-efficient inference with MoE architecture
Sources

Source: https://huggingface.co/google/gemma-4-26B-A4B-it