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...

Modalities
ImageTextVideo
Giá In / out
123 ₩ / 597 ₩
1M
Context
262,14 N
Ngày phát hành
3 thg 4, 2026

Apps

Các ứng dụng sử dụng model này nhiều nhất.

Appstokens
Hermes Agent11,32 NT
Zazen (Freebuff fork)3,16 NT
Cline2,56 NT
Kilo Code2,44 NT
pi1,59 NT
OpenClaw1,05 NT
Codex938,81 T
DeepSeek Harness847,26 T
Portkey AI639,39 T
HighLevel375,1 T
ISEKAI ZERO364,56 T
Framer348,63 T
Hello Minds, powered by Ethoswarm316,97 T
ZCode279,83 T
Descript253,44 T