zhamm/ qwopus3.6

59 Downloads Updated 1 week ago

Qwopus3.6-27B-v2-MTP-Q8 Vision - A custom import of Qwopus3.6-27B-v2-MTP.GGUF model with multimodal vision projector support quantized at Q8.

vision

ollama run zhamm/qwopus3.6:27b-q8-vision

curl http://localhost:11434/api/chat \
  -d '{
    "model": "zhamm/qwopus3.6:27b-q8-vision",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='zhamm/qwopus3.6:27b-q8-vision',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'zhamm/qwopus3.6:27b-q8-vision',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size / Usage

Context

Input

qwopus3.6:27b-q8-vision

30GB · 256K context window · Text, Image · 1 week ago

qwopus3.6:27b-q8-vision

30GB

256K

Text, Image

Readme

Qwopus3.6 27B v2 MTP Q8 Vision

Custom Ollama import of the Qwopus3.6 27B v2 MTP GGUF model with multimodal projector support.

Model Details

Base model: Qwopus3.6-27B-v2-MTP
Quantization: Q8_0
Vision projector: mmproj-F32.gguf
Approximate model size: 29 GB
Approximate projector size: 0.93 GB
Approximate total size: 30 GB

Intended Use

This model is intended for local AI use with Ollama, including:

General chat
Reasoning
Coding assistance
Document analysis
Image-text tasks where supported by the Ollama runtime and client

Recommended Hardware

Recommended GPU VRAM:

Minimum: 48 GB VRAM
Preferred: 64 GB+ VRAM
Ideal: 96 GB VRAM for larger context windows and experimentation

Smaller GPUs may require reduced context length or may fall back to CPU/RAM offload, which can be much slower.

Suggested Settings

Recommended starting settings:

Context length: 32768
Temperature: 0.7
Top-p: 0.9

For long-context use, Flash Attention and q8_0 KV cache are recommended where supported.

Usage

ollama run <your-username>/<model-name>

Notes

This build includes a multimodal projector for image-text use. Vision support depends on the Ollama runtime, the client being used, and how image input is passed to the model.

If image input does not work through a specific UI, test directly with the Ollama API or use a runtime with explicit mmproj support.

Attribution

Source model: Jackrong/Qwopus3.6-27B-v2-MTP-GGUF on Hugging Face.

Please review and follow the upstream model license before redistribution or commercial use.

# Qwopus3.6 27B v2 MTP Q8 Vision

Custom Ollama import of the Qwopus3.6 27B v2 MTP GGUF model with multimodal projector support.

## Model Details

* Base model: `Qwopus3.6-27B-v2-MTP`
* Quantization: `Q8_0`
* Vision projector: `mmproj-F32.gguf`
* Approximate model size: 29 GB
* Approximate projector size: 0.93 GB
* Approximate total size: 30 GB

## Intended Use

This model is intended for local AI use with Ollama, including:

* General chat
* Reasoning
* Coding assistance
* Document analysis
* Image-text tasks where supported by the Ollama runtime and client

## Recommended Hardware

Recommended GPU VRAM:

* Minimum: 48 GB VRAM
* Preferred: 64 GB+ VRAM
* Ideal: 96 GB VRAM for larger context windows and experimentation

Smaller GPUs may require reduced context length or may fall back to CPU/RAM offload, which can be much slower.

## Suggested Settings

Recommended starting settings:

```text
Context length: 32768
Temperature: 0.7
Top-p: 0.9
```

For long-context use, Flash Attention and `q8_0` KV cache are recommended where supported.

## Usage

```bash
ollama run <your-username>/<model-name>
```

## Notes

This build includes a multimodal projector for image-text use. Vision support depends on the Ollama runtime, the client being used, and how image input is passed to the model.

If image input does not work through a specific UI, test directly with the Ollama API or use a runtime with explicit `mmproj` support.

## Attribution

Source model: `Jackrong/Qwopus3.6-27B-v2-MTP-GGUF` on Hugging Face.

Please review and follow the upstream model license before redistribution or commercial use.

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)