4-bit K-quntized version of google/gemma-4-E4B-it. Generated using llama.cpp

tools thinking

ollama run su_robin/gemma-4-E4B-it-Q4_K_M

curl http://localhost:11434/api/chat \
  -d '{
    "model": "su_robin/gemma-4-E4B-it-Q4_K_M",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='su_robin/gemma-4-E4B-it-Q4_K_M',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'su_robin/gemma-4-E4B-it-Q4_K_M',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 4 months ago

4 months ago

75c5c8502945 · 5.3GB ·

model

archgemma4

parameters7.52B

quantizationQ4_K_M

5.3GB

params

{ "stop": [ "<turn|>" ] }

31B

Readme

Gemma-4-E4B-it-Q4_K_M

Lightweight redistribution of the Gemma-4-E4B-it instruction-tuned model. Conveted to gguf format with 4-bit K-Quantization (Q4_K_M) using llama.cpp.

Attribution

Model source: Hugging Face — https://huggingface.co/google/gemma-4-E4B-it
Quantization: llama.cpp — https://github.com/ggerganov/llama.cpp

All credit belongs to the original authors and contributors.

Lisence

Model weights from Google Deepmind under the Apache 2.0 Lisence