mannix/ z-image

194 Downloads Updated 1 week ago

Diffusion model · Qwen3-4B text encoder · 1024×1024 in eight steps

ollama run mannix/z-image:turbo

curl http://localhost:11434/api/chat \
  -d '{
    "model": "mannix/z-image:turbo",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='mannix/z-image:turbo',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'mannix/z-image:turbo',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size / Usage

Context

Input

z-image:turbo

6.4GB · - context window · Text · 1 week ago

z-image:turbo

6.4GB

-

Text

Readme

Z-Image Turbo — photographic image generation, for xOllama

Diffusion model (Q4_0 GGUF) · Qwen3-4B text encoder · 1024×1024 in eight steps · OpenAI /v1/images routes

Runs on xOllama, not on stock ollama. Stock ollama pulls this model but cannot show or run it: it carries only media engines. xOllama serves it through an opencoti engine with media support (dev build 2610022033001/b105 or later, set with XOLLAMA_ENGINE_PATH). The engine pinned by the current xOllama release has no media engines and refuses the model at load, naming what it lacks.

Z-Image Turbo is the photographic one: more detail than FLUX.2 klein, in eight steps. It shares klein’s text encoder, so a server with both stores it once. Edits are img2img.

Use

xollama pull mannix/z-image:turbo
curl http://localhost:22434/v1/images/generations \
  -d '{"model": "mannix/z-image:turbo", "prompt": "a fishing harbour at 6 a.m., 35 mm photograph"}'

Any OpenAI client works: the reply is b64_json, or a data: URL with "response_format": "url". Besides OpenAI’s fields the engine takes steps, cfg_scale, sampler, scheduler, seed and negative_prompt.

What the template fills in

A client that sends only a prompt gets 1024×1024, 8 steps, cfg 1, png.

Measured

	RTX 3090
1024×1024 generation	36 s (opencoti b97)

Tags

Tag	Weights	Size
`turbo`	`z_image_turbo-Q4_0.gguf` + Qwen3-4B Q4_0 + VAE	6.4 GB

Source

GGUF: leejet/Z-Image-Turbo-GGUF · VAE: Comfy-Org/z_image_turbo · text encoder: unsloth/Qwen3-4B-GGUF · License: Apache-2.0. Docs: xOllama media.

# Z-Image Turbo — photographic image generation, for xOllama

**Diffusion model (Q4_0 GGUF) · Qwen3-4B text encoder · 1024×1024 in eight steps · OpenAI `/v1/images` routes**

> **Runs on [xOllama](https://github.com/mann1x/xollama), not on stock ollama.** Stock ollama pulls this model but cannot show or run it: it carries only media engines. xOllama serves it through an [opencoti](https://huggingface.co/ManniX-ITA/opencoti-llamafile) engine with media support (dev build `2610022033001`/b105 or later, set with `XOLLAMA_ENGINE_PATH`). The engine pinned by the current xOllama release has no media engines and refuses the model at load, naming what it lacks.

Z-Image Turbo is the photographic one: more detail than FLUX.2 klein, in eight steps. It shares klein's text encoder, so a server with both stores it once. Edits are img2img.

## Use

```shell
xollama pull mannix/z-image:turbo
curl http://localhost:22434/v1/images/generations \
  -d '{"model": "mannix/z-image:turbo", "prompt": "a fishing harbour at 6 a.m., 35 mm photograph"}'
```

Any OpenAI client works: the reply is `b64_json`, or a `data:` URL with `"response_format": "url"`. Besides OpenAI's fields the engine takes `steps`, `cfg_scale`, `sampler`, `scheduler`, `seed` and `negative_prompt`.

## What the template fills in

A client that sends only a prompt gets **1024×1024, 8 steps, cfg 1, png**.

## Measured

| | RTX 3090 |
|---|---|
| 1024×1024 generation | 36 s (opencoti b97) |

## Tags

| Tag | Weights | Size |
|---|---|---|
| `turbo` | `z_image_turbo-Q4_0.gguf` + Qwen3-4B Q4_0 + VAE | 6.4 GB |

## Source

GGUF: [leejet/Z-Image-Turbo-GGUF](https://huggingface.co/leejet/Z-Image-Turbo-GGUF) · VAE: [Comfy-Org/z_image_turbo](https://huggingface.co/Comfy-Org/z_image_turbo) · text encoder: [unsloth/Qwen3-4B-GGUF](https://huggingface.co/unsloth/Qwen3-4B-GGUF) · License: Apache-2.0.
Docs: [xOllama media](https://github.com/mann1x/xollama/blob/dev/docs/xollama/media.mdx).

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