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
ImagemTextoVídeo
Preço in / out
₩ 123 / ₩ 597
1M
Contexto
262,14 mil
Lançado
3 de abr. de 2026
Apps
Apps que mais utilizam este modelo.
| Apps | tokens |
|---|---|
| Hermes Agent | 11,32 tri |
| Zazen (Freebuff fork) | 3,16 tri |
| Cline | 2,56 tri |
| Kilo Code | 2,44 tri |
| pi | 1,59 tri |
| OpenClaw | 1,05 tri |
| Codex | 938,81 bi |
| DeepSeek Harness | 847,26 bi |
| Portkey AI | 639,39 bi |
| HighLevel | 375,1 bi |
| ISEKAI ZERO | 364,56 bi |
| Framer | 348,63 bi |
| Hello Minds, powered by Ethoswarm | 316,97 bi |
| ZCode | 279,83 bi |
| Descript | 253,44 bi |