Instructions to use pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit 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("pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit")
config = load_config("pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit")

# 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 pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit"

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": "pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit 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 "pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit"

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 pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit

Run Hermes

hermes

OpenClaw new

How to use pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit"

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 "pipenetwork/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit" \
  --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"

pipenetwork commited on 8 days ago

Commit

fb05bcd

verified ·

1 Parent(s): af5dfec

Benchmark all quant tiers against the bf16 reference

Browse files

Files changed (1) hide show

README.md +24 -14

README.md CHANGED Viewed

@@ -26,7 +26,7 @@ base_model: nightmedia/Qwen3.5-9B-DS9-USS-Defiant
 # Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit
-MLX conversion of **Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic** for Apple silicon — 4-bit, the smallest tier and the usual default
 Uncensored ("Heretic'd") multi-stage merge of Qwen3.5-9B fine tunes by [nightmedia](https://huggingface.co/nightmedia) and [DavidAU](https://huggingface.co/DavidAU), with a compacted-but-stronger thinking block. Vision is included and works out of the box — no separate `mmproj` download.
@@ -108,19 +108,29 @@ Group size 64, `affine` mode. The vision tower is left unquantized (mlx-vlm's de
 ## Measured quantization quality
 Against the bf16 reference these were converted from — 65,536 tokens of wikitext-2 test
-at 1024 context, identical token ids through mlx-lm, on an M3 Ultra:
-| model | ppl | Δppl | KL(bf16‖quant) | top-1 vs bf16 | decode |
-|---|---|---|---|---|---|
-| bf16 reference | 8.1273 | — | — | — | 39.0 t/s |
-| **this 4bit** (affine, group 64) | **8.5816** | **+5.59%** | **0.0733** | **87.06%** | 109.9 t/s |
-| [nightmedia mxfp4](https://huggingface.co/nightmedia/Qwen3.5-9B-DS9-USS-Defiant-mxfp4-mlx) (group 32) | 8.9443 | +10.05% | 0.1133 | 82.34% | 114.4 t/s |
-The 4bit tier loses about half the perplexity that MXFP4 does and stays measurably closer
-to the original distribution, at the cost of 4.5 vs 4.25 bits/weight (~6% more storage)
-and ~4% slower decode. Both are quantizations of the same weights, so this compares the
-quantization scheme only. Reproduce with
-[`bench.py`](https://github.com/PipeNetwork/defiant-fable-mlx).
 ## Sampling

 # Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-MLX-4bit
+MLX conversion of **Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic** for Apple silicon — 4-bit, the smallest tier
 Uncensored ("Heretic'd") multi-stage merge of Qwen3.5-9B fine tunes by [nightmedia](https://huggingface.co/nightmedia) and [DavidAU](https://huggingface.co/DavidAU), with a compacted-but-stronger thinking block. Vision is included and works out of the box — no separate `mmproj` download.
 ## Measured quantization quality
 Against the bf16 reference these were converted from — 65,536 tokens of wikitext-2 test
+at 1024 context, identical token ids through mlx-lm, on an M3 Ultra. KL is measured
+against bf16's own output distribution, so lower means closer to the original model.
+| model | lang. weights | ppl | Δppl | KL(bf16‖quant) | top-1 vs bf16 | decode |
+|---|---|---|---|---|---|---|
+| bf16 reference | 17.91 GB | 8.1273 | — | — | — | 38.9 t/s |
+| 8bit | 9.51 GB | 8.1277 | +0.00% | 0.00124 | 98.24% | 67.3 t/s |
+| 6bit | 7.28 GB | 8.1426 | +0.19% | 0.00523 | 96.20% | 80.4 t/s |
+| **4bit** (this repo) | 5.04 GB | 8.5816 | +5.59% | 0.07330 | 87.06% | 109.1 t/s |
+| [nightmedia mxfp4](https://huggingface.co/nightmedia/Qwen3.5-9B-DS9-USS-Defiant-mxfp4-mlx) | 4.76 GB | 8.9443 | +10.05% | 0.11328 | 82.34% | 113.8 t/s |
+Reading it:
+- **8bit is effectively free** — bf16 perplexity to four decimals at 47% of the footprint
+  and 1.7x the decode speed.
+- **6bit is the value pick** at +0.19%.
+- **The quality cliff is 6bit → 4bit**, not further down: KL jumps 14x. If you can fit
+  7.3 GB, prefer 6bit over 4bit.
+- At the 4-bit tier this affine group-64 quant loses about half the perplexity MXFP4 does,
+  costing 4.5 vs 4.25 bits/weight (~6% more storage, ~4% slower decode).
+All tiers are quantizations of the same weights, so this isolates the quantization scheme.
+Reproduce with [`bench.py`](https://github.com/PipeNetwork/defiant-fable-mlx).
 ## Sampling