Instructions to use OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4 with MLX:

# Make sure mlx-vlm is installed
# pip install --upgrade mlx-vlm

from mlx_vlm import load, generate
from mlx_vlm.prompt_utils import apply_chat_template
from mlx_vlm.utils import load_config

# Load the model
model, processor = load("OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4")
config = load_config("OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4")

# Prepare input
image = ["http://images.cocodataset.org/val2017/000000039769.jpg"]
prompt = "Describe this image."

# Apply chat template
formatted_prompt = apply_chat_template(
    processor, config, prompt, num_images=1
)

# Generate output
output = generate(model, processor, formatted_prompt, image)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4 with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4"

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4

Run Hermes

hermes

OpenClaw new

How to use OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4 with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4"

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "OsaurusAI/diffusiongemma-26B-A4B-it-MXFP4" \
  --custom-provider-id mlx-lm \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

diffusiongemma-26B-A4B-it-MXFP4 / README.md

Osaurus-AI

Add DiffusionGemma MXFP4 bundle (vmlx block-diffusion engine)

d523a32 verified about 1 month ago

preview code

Raw

History Blame Contribute Delete

2.35 kB

	---
	license: gemma
	base_model: google/diffusiongemma-26B-A4B-it
	pipeline_tag: image-text-to-text
	tags:
	- mlx
	- vmlx
	- osaurus
	- diffusion-language-model
	- block-diffusion
	- mxfp4
	- gemma
	library_name: mlx
	---

	![Osaurus](osaurus-x-banner.png)

	# DiffusionGemma 26B-A4B-it — MXFP4 (Osaurus / vMLX)

	Native MLX MXFP4 quantization of `google/diffusiongemma-26B-A4B-it` — a
	block-diffusion language model (NOT autoregressive): text generates as
	256-token canvases refined by iterative denoising. 30-layer Gemma-4-style
	MoE, 128 experts top-8, 26B total / ~4B active parameters.

	Runs natively in [Osaurus](https://github.com/osaurus-ai/osaurus) on Apple
	Silicon via the vmlx-swift block-diffusion engine.

	## Quantization

	- Attention + routed MoE experts: MXFP4 (group 32)
	- Dense MLP + router projections: MXFP8 (group 32)
	- Embeddings, norms, self-conditioning, vision tower: fp16 passthrough
	- 15 shards, ~15 GB on disk, peak runtime RSS ≈ 12.7 GB (M5 Max)

	## Capabilities

	\| \| \|
	\|---\|---\|
	\| Text generation \| ✅ block diffusion (~37 tok/s @ bundle default 48 steps, ~74 tok/s @ 16 steps, M5 Max) \|
	\| Vision (single/multi image) \| ✅ Gemma-4 unified vision tower, 280 soft tokens/image \|
	\| Tool calling \| ✅ Gemma-4 format `<\\|tool_call>call:name{...}<tool_call\\|>` \|
	\| Reasoning channel \| ✅ harmony `<\\|channel>thought…<channel\\|>` \|
	\| Audio \| ❌ not in this checkpoint (no audio_config) \|
	\| Video \| ❌ no `video_token_id` \|

	## Generation contract

	All diffusion sampling parameters live in `generation_config.json` and are
	honored by the runtime: `max_denoising_steps=48`, entropy bound 0.1,
	temperature schedule 0.8→0.4, stability 1, confidence 0.005,
	`eos_token_id=[1, 106, 50]`, pad 0. Wire `temperature`/`top_p` are ignored
	by design — the denoising schedule is bundle-owned. Speed/quality is
	controlled by the denoising-step budget (Osaurus exposes this as a server
	setting, default 16 ≈ 2× faster than the bundle default and verified
	coherent; below 12 quality degrades).

	The chat template (`chat_template.jinja`) ships in this repo, including
	tool-call and thinking-channel rendering.

	## Known behavior

	Very terse prompts under greedy denoising can occasionally converge to an
	empty (EOS-first) canvas — inherent to the reference sampling algorithm
	with random canvas initialization; retry or rephrase.