second_constantine/
gemma3-qat:27b

89 1 month ago

Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. (quantized IQ4_XS)

vision 27b

1 month ago

43dea44ca2ec · 16GB ·

model
gemma3
·
27B
·
IQ4_XS
projector
clip
·
423M
·
F16
template
{{- $systemPromptAdded := false }} {{- range $i, $_ := .Messages }} {{- $last := eq (len (slice $.Me
params
{ "stop": [ "<end_of_turn>" ], "top_k": 64, "top_p": 0.95 }

Readme

Feature Value
vision true (>=0.11.11)
thinking false
tools false
Device Speed, token/s Context VRAM, gb Versions
RTX 3090 24gb ~30 4096 17 IQ4_XS,0.12.2
RTX 3090 24gb ~30 15360 19 IQ4_XS,0.12.2
M1 Max 32gb ~14 4096 17 IQ4_XS,0.12.2
M1 Max 32gb ~13 15360 18 IQ4_XS,0.12.2