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...
| Type | Model |
| Section | Models & platforms / text+image+video->text |
| Pricing | paid (от $0.07/mo) |
| Platform | Self-hosted |
| Systems | api, python, self-hosted |
| Hosting | cloud |
| Site language | en |
| Vendor | |
| Launched | 2026-04-03 |