SOURCE-LINKED INTELLIGENCE
Google: Gemma 4 26B A4B (free)
Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token during inference — delivering near-31B quality at...
Read original source ↗ Open in workspace
- recordType
- model-catalog
- evidenceStatus
- provider-reported
- region
- Global
Evidence & attribution
First collected: 2026-09-20T08:01:03.945Z. This is not the publication date.