Gemma in an office printer? @osanseviero says yes—at one minute per token
@osanseviero reports running quantized Gemma 3 270m as one PostScript job inside a printer, with 14MB peak RAM use.
TLDR
@osanseviero says Carbon helped run quantized Gemma 3 270m entirely inside an office printer as a single PostScript job. Reported peak RAM use was 14MB, at one minute per token of generated text. At that pace, patience is part of the setup. We haven’t independently verified the setup or performance.
Combined views
6.4K
2 Sources, first seen ago

Comments