frozenlab/gemma4-mtp:12b

frozenlab/ gemma4-mtp:12b

164 Downloads Updated 1 month ago

vision tools thinking 12b

ollama run frozenlab/gemma4-mtp:12b

curl http://localhost:11434/api/chat \
  -d '{
    "model": "frozenlab/gemma4-mtp:12b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='frozenlab/gemma4-mtp:12b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'frozenlab/gemma4-mtp:12b',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 1 month ago

1 month ago

d36888b00170 · 8.0GB

model

archgemma4

parameters11.9B

quantizationQ4_K_M

7.4GB

projector

archclip

parameters52.4M

quantizationBF16

175MB

license

Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR US

10kB

draft

archgemma4-assistant

parameters423M

quantizationQ8_0

465MB

params

{ "draft_num_predict": 4, "temperature": 1, "top_k": 64, "top_p": 0.95 }

64B

Readme

Base model: Gemma 4 12B instruction-tuned (Q4_K_M), identical to gemma4:12b
Drafter: Google’s gemma-4-12B-it-assistant multi-token-prediction head, Q8_0 GGUF by ggml-org (ggml-org/gemma-4-12B-it-GGUF)
Parameter: draft_num_predict 4

The weights are unmodified. This package only combines the two published models into a single Ollama model. Vision, chat template and sampling defaults are inherited from gemma4:12b.

Requirements

Ollama 0.32.7 or newer. Older versions can pull and run the model but silently skip the drafter, so you get no speedup.

License

Apache License 2.0. The license text ships with the model as a separate layer. Gemma is a trademark of Google LLC. This is not an official Google model and is not affiliated with or endorsed by Google.

Details

Readme

Contents

Requirements

License