Instructions to use zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 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("zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8")
config = load_config("zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8")

# 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)

Transformers

How to use zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8")
model = AutoModelForMultimodalLM.from_pretrained("zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

vLLM

How to use zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8

SGLang

How to use zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Unsloth Studio

How to use zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8",
    max_seq_length=2048,
)

How to use zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8"

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": "zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 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 "zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8"

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 zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8

Run Hermes

hermes

Docker Model Runner
How to use zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8 with Docker Model Runner:
```
docker model run hf.co/zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8
```

This doesnt work on open LLM some error with Vision

by ralong - opened Apr 25

Discussion

ralong

Apr 25

🥲 Failed to load the model

Failed to load model.

Error when loading model: ValueError: Missing 1 parameters: 
embed_vision.embedding_projection.biases.

2026-04-26 01:06:06 [DEBUG]
The tokenizer you are loading from '/Users/oskar/.lmstudio/models/zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the fix_mistral_regex=True flag when loading this tokenizer to fix this issue.
2026-04-26 01:06:06 [DEBUG]
[model_kit][INFO]: Loading model from /Users/oskar/.lmstudio/models/zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8...
2026-04-26 01:06:14 [DEBUG]
The tokenizer you are loading from '/Users/oskar/.lmstudio/models/zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the fix_mistral_regex=True flag when loading this tokenizer to fix this issue.
2026-04-26 01:06:16 [DEBUG]
The tokenizer you are loading from '/Users/oskar/.lmstudio/models/zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-mxfp8' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the fix_mistral_regex=True flag when loading this tokenizer to fix this issue.
2026-04-26 01:06:16 [DEBUG]
/Users/oskar/.lmstudio/extensions/backends/vendor/_amphibian/app-mlx-generate-mac14-arm64@22/lib/python3.11/site-packages/transformers/audio_utils.py:538: UserWarning: At least one mel filter has all zero values. The value for num_mel_filters (128) may be set too high. Or, the value for num_frequency_bins (257) may be set too low.
warnings.warn(
2026-04-26 01:06:17 [DEBUG]
ValueError: Missing 1 parameters:
embed_vision.embedding_projection.biases.

At:
/Users/oskar/.lmstudio/extensions/backends/vendor/_amphibian/app-mlx-generate-mac14-arm64@22/lib/python3.11/site-packages/mlx/nn/layers/base.py(191): load_weights
/Users/oskar/.lmstudio/extensions/backends/vendor/_amphibian/app-mlx-generate-mac14-arm64@22/lib/python3.11/site-packages/mlx_engine/model_kit/vision_add_ons/load_utils.py(182): prepare_components
/Users/oskar/.lmstudio/extensions/backends/vendor/_amphibian/app-mlx-generate-mac14-arm64@22/lib/python3.11/site-packages/mlx_engine/model_kit/vision_add_ons/gemma4.py(71): init
/Users/oskar/.lmstudio/extensions/backends/vendor/_amphibian/app-mlx-generate-mac14-arm64@22/lib/python3.11/site-packages/mlx_engine/model_kit/model_kit.py(120): _full_model_init
/Users/oskar/.lmstudio/extensions/backends/vendor/_amphibian/app-mlx-generate-mac14-arm64@22/lib/python3.11/site-packages/mlx_engine/model_kit/model_kit.py(141): init
/Users/oskar/.lmstudio/extensions/backends/vendor/_amphibian/app-mlx-generate-mac14-arm64@22/lib/python3.11/site-packages/mlx_engine/generate.py(253): load_model
2026-04-26 01:06:17 [DEBUG]
[LLMProcess] Failed to load model _0x5b3431 [Error]: Failed to load model.
at _0x1186d7.loadModel (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:1:610408)
at process.processTicksAndRejections (node:internal/process/task_queues:104:5)
at async _0x1186d7.handleMessage (/Applications/LM Studio.app/Contents/Resources/app/.webpack/lib/llmworker.js:1:602598) {
cause: 'Error when loading model: ValueError: Missing 1 parameters: \n' +
'embed_vision.embedding_projection.biases.',
suggestion: undefined,
errorData: undefined,
data: undefined,
displayData: undefined,
title: 'Failed to load model.'
}
2026-04-26 01:06:17 [DEBUG]
stopGenerating() without a request_id is deprecated. Taking no action

zecanard

Owner Apr 26

Unfortunately there seems to be a bug in the MLX engine with MXFP builds right now. You can try one of the affine quants in the meantime:
https://huggingface.co/zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-int8-affine

ralong

Apr 26

Will there be a v3 fix for that?

Unfortunately there seems to be a bug in the MLX engine with MXFP builds right now. You can try one of the affine quants in the meantime:
https://huggingface.co/zecanard/gemma-4-26B-A4B-it-Claude-Opus-Distilled-v2-MLX-8bit-int8-affine

zecanard

Owner Apr 26

There shouldn’t be a need for a new quant since it is strictly a bug in the MLX engine. Once the fix goes in, you can try this conversion again to see if it works.

If TeichAI puts out a v3 of this finetune and I don’t notice it, just let me know and I’ll be happy to convert it to MLX.

ralong

Apr 26

So the fix will be in the LM studio itself not in the model ?

There shouldn’t be a need for a new quant since it is strictly a bug in the MLX engine. Once the fix goes in, you can try this conversion again to see if it works.

If TeichAI puts out a v3 of this finetune and I don’t notice it, just let me know and I’ll be happy to convert it to MLX.

zecanard

Owner Apr 26

It will be in the MLX engine.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment