Clef MLX

MLX quantizations of Cloudflare/clef, a 27B multimodal decision model post-trained from Qwen3.8-27B.

MLX Files

Usage with MLX-VLM

Installation

pip install -U mlx-vlm huggingface_hub

Python API

Download the target quantization and load directly with MLX-VLM:

from huggingface_hub import snapshot_download
from mlx_vlm import load, generate
from mlx_vlm.prompt_utils import apply_chat_template

# 1. Download target quantization (Clef-MLX-4bit, Clef-MLX-6bit, or Clef-MLX-8bit)
subfolder = "Clef-MLX-4bit"
model_dir = snapshot_download("abenzerps/Clef-MLX", allow_patterns=f"{subfolder}/*")
model_path = f"{model_dir}/{subfolder}"

# 2. Load model and processor
model, processor = load(model_path)
config = model.config

# Text prompt
prompt = "Explain why reproducible builds matter."
formatted_prompt = apply_chat_template(processor, config, prompt)

output = generate(model, processor, formatted_prompt, verbose=True)
print(output)

For image input:

prompt = "Describe this image."
image = ["image.jpg"]

formatted_prompt = apply_chat_template(
    processor, config, prompt, num_images=len(image)
)

output = generate(model, processor, formatted_prompt, image, verbose=True)
print(output)

Command Line Interface

Download the quantization folder and run inference via CLI:

# Download 4-bit quantization folder
huggingface-cli download abenzerps/Clef-MLX --include "Clef-MLX-4bit/*" --local-dir ./Clef-MLX

# 4-bit Text
python -m mlx_vlm.generate \
    --model ./Clef-MLX/Clef-MLX-4bit \
    --prompt "Explain why reproducible builds matter."

# 4-bit Image
python -m mlx_vlm.generate \
    --model ./Clef-MLX/Clef-MLX-4bit \
    --image image.jpg \
    --prompt "Describe this image."

Original Python / Transformers Usage

Tested with torch 2.11 and transformers 5.10.2 on a single H200. Image and video inputs also need pillow.

import sys

import torch
from huggingface_hub import snapshot_download

path = snapshot_download("Cloudflare/clef")
sys.path.insert(0, path)
from joint_schema_model import collate_records, encode_record, load_release_model

model, processor = load_release_model(path, device="cuda")

record = {
    "state": {"invoice": {"vendor": "Acme", "total": 1250.0, "currency": "USD", "status": "overdue"}},
    "questions": {
        "status": {
            "type": "choice",
            "instructions": "What is the invoice status?",
            "criteria": {"paid": "Invoice is paid.", "overdue": "Invoice is past due.", "draft": "Not sent."},
        },
        "large": {"type": "noul", "instructions": "Is the total above 1000 USD?"},
    },
}

encoded = encode_record(processor.tokenizer, record, processor=processor)
batch = collate_records([encoded], processor.tokenizer.pad_token_id, torch.device("cuda"))
with torch.inference_mode():
    logits = model(batch)[0]

for question, question_logits in zip(encoded.questions, logits):
    probabilities = question_logits.float().softmax(-1).tolist()
    print(question.question_id, dict(zip(question.option_ids, probabilities)))

Jev / SystemOne API

systemone takes a Jev/SystemOne POST /v1/systemone request body and returns the same response body: model, answers keyed by question ID, and usage. A choice answer has choice, confidence, and probabilities; a score answer has the expected score, confidence, legend, and probabilities; a noul answer has the probability of true. instructions is optional, and images and videos may be added to the request.

from joint_schema_model import systemone

response = systemone(model, processor, {
    "model": "clef",
    "state": "Our checkout started returning errors and orders are blocked.",
    "questions": {
        "department": {
            "type": "choice",
            "instructions": "Which team should handle the message?",
            "criteria": {"billing": "Payments or invoices", "technical": "Bugs or outages"},
        },
        "urgency": {"type": "score", "criteria": ["Can wait", "This week", "Today"]},
        "outage": {"type": "noul", "instructions": "Is a service down?"},
    },
})
print(response["answers"])

Images and video

Add images (PIL images) or videos (frame arrays) to the record and pass the processor to encode_record. Optional processor arguments go in media_kwargs.

from PIL import Image

record = {
    "state": {"task": "Review the attached receipt."},
    "images": [Image.open("receipt.jpg")],
    "questions": {
        "legible": {"type": "noul", "instructions": "Is the receipt total legible?"},
    },
}
encoded = encode_record(processor.tokenizer, record, processor=processor)

Text-only and multimodal records can be mixed in the same batch.

Each question has:

• type: noul (true/false), choice (named options), or score (ordered options)
• instructions: what to decide; optional, and the question ID is used when it is omitted
• criteria: for choice, a mapping of option ID to description; for score, a list of option descriptions indexed from 0; for noul, optional descriptions for true and false

encode_record accepts max_length (default 16,384 tokens) and max_state_tokens to bound the input.

Benchmarks

Decision Index & Workflow Evals

Decision Index

Per-benchmark results from our internal run of the Decision Index 0.2.1 suite. Scores are percentages; ForecastBench is a Brier score, where lower is better. The last two rows are request latency in milliseconds, where lower is better. The best value in each row is in bold.

Workflow evals

Decision accuracy on four end-to-end business workflows from Typesafe Evals, scored against consensus reference labels. All models are scored on the same dataset revision and case cohort.

Source

• Model: Cloudflare/clef
• Base model: Qwen/Qwen3.8-27B
• License: Apache-2.0
• Checksums: SHA256SUMS