Google: Gemma 4 26B A4B
google/gemma-4-26b-a4b-it
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...
Modalidades
ImagenTextoVideo
Precio entrada / salida
123 KRW / 597 KRW
1M
Contexto
262,14 mil
Lanzado
3 abr 2026
Apps
Apps que más utilizan este modelo.
| Apps | tokens |
|---|---|
| Hermes Agent | 11,32 B |
| Zazen (Freebuff fork) | 3,16 B |
| Cline | 2,56 B |
| Kilo Code | 2,44 B |
| pi | 1,59 B |
| OpenClaw | 1,05 B |
| Codex | 938,81 mil M |
| DeepSeek Harness | 847,26 mil M |
| Portkey AI | 639,39 mil M |
| HighLevel | 375,1 mil M |
| ISEKAI ZERO | 364,56 mil M |
| Framer | 348,63 mil M |
| Hello Minds, powered by Ethoswarm | 316,97 mil M |
| ZCode | 279,83 mil M |
| Descript | 253,44 mil M |