OpenWeightsTerminal
Models

Gemma-4 26B-A4B-it

Google · MoE

Official QAT line. Street UD-Q4_K_XL llama.cpp quote on one RX 9060 XT.

Params

26B-A4B

Fastest decode

88 tok/s

Radeon RX 9060 XT 16GB · UD-Q4_K_XL · llama.cpp v0.6.0

Refusal

Official

Setups

1

Best reported pair · not live serving speed ↗

Run this model

Runs on

Quoted one-stream decode on real boxes.

View all

Get it running

  • 1. DownloadPuts the weights on this computer.
    huggingface-cli download google/gemma-4-26b-a4b-it
  • 2. Start itUse this only if the download is GGUF.
    llama-server -hf google/gemma-4-26b-a4b-it
  • Open the weightsHugging Face ↗