0
MODEL SIGNAL · GOOGLE

gemma-4-26B-A4B-it

A 26-billion-parameter, instruction-tuned Mixture-of-Experts model from Google supporting multimodal inputs and a 256K-token context window.

CATEGORYMultimodal
CONTEXT256K
RELEASEDApril 3, 2026
Key Features
  • Mixture-of-Experts (MoE) architecture with 25.2B total and 3.8B active parameters
  • Multimodal input support for text, image, and video
  • 256K-token context window
  • Instruction-tuned variant
  • Open weights released under Apache 2.0 license

Provider announcement →

Read the Model Signal report →

The designed Model Signal report for gemma-4-26B-A4B-it is still in review. It will publish here after the fact and quality gates clear.