Image-Text-to-Text
MLX
Safetensors
qwen3_vl
apple-silicon
quantized
mixed-precision
axquant
axq
development
qwen3-vl
MXFP4
vision
conversational
4-bit precision
Instructions to use AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-MXFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-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("AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-MXFP4") config = load_config("AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-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
- Pi
How to use AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-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 "AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-MXFP4"
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "mlx-lm": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-MXFP4" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Hermes Agent
How to use AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-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 "AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-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 AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-MXFP4
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-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 "AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-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 "AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-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"
Record architecture-specific runtime format audit and n-gram scope
Browse files- README.md +11 -0
- axquant_manifest.json +8 -3
- runtime_audit.json +51 -0
README.md
CHANGED
|
@@ -18,6 +18,17 @@ tags:
|
|
| 18 |
- vision
|
| 19 |
---
|
| 20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
# AX-Qwen3-VL-4B-Instruct-MLX-AXQ-MXFP4 — 5.72 BPW measured main
|
| 22 |
|
| 23 |
An **AXQuant (AXQ)** mixed-precision MLX checkpoint for Apple Silicon, converted directly from
|
|
|
|
| 18 |
- vision
|
| 19 |
---
|
| 20 |
|
| 21 |
+
## Runtime format audit (2026-10-06)
|
| 22 |
+
|
| 23 |
+
No quantization-container correction was needed.
|
| 24 |
+
This family has no n-gram tensors; no n-gram file or declaration was added.
|
| 25 |
+
This remote format audit does not grant an oMLX/MTPLX runtime profile or a successful load/generation claim.
|
| 26 |
+
|
| 27 |
+
See [runtime_audit.json](runtime_audit.json) for pinned config/index/header
|
| 28 |
+
bindings, architecture, physical-format findings, and applied corrections.
|
| 29 |
+
This is development evidence; no quality, MTP exactness, speed, or certification
|
| 30 |
+
claim is added. Historical evidence stays bound to its original revision.
|
| 31 |
+
|
| 32 |
# AX-Qwen3-VL-4B-Instruct-MLX-AXQ-MXFP4 — 5.72 BPW measured main
|
| 33 |
|
| 34 |
An **AXQuant (AXQ)** mixed-precision MLX checkpoint for Apple Silicon, converted directly from
|
axquant_manifest.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
"axquant_version": "1.9.0",
|
| 3 |
"calibration": null,
|
| 4 |
-
"created_at": "2026-10-
|
| 5 |
"effective_bpw": 6.336767613767534,
|
| 6 |
"files": [
|
| 7 |
{
|
|
@@ -161,8 +161,8 @@
|
|
| 161 |
},
|
| 162 |
{
|
| 163 |
"path": "README.md",
|
| 164 |
-
"sha256": "
|
| 165 |
-
"size_bytes":
|
| 166 |
},
|
| 167 |
{
|
| 168 |
"path": "axquant_plan.json",
|
|
@@ -229,6 +229,11 @@
|
|
| 229 |
"sha256": "45fc17c8dd2474af6b493b52483c26c0584b0082d368c480f9fa611e73070040",
|
| 230 |
"size_bytes": 991
|
| 231 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 232 |
{
|
| 233 |
"path": "tokenizer.json",
|
| 234 |
"sha256": "be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506",
|
|
|
|
| 1 |
{
|
| 2 |
"axquant_version": "1.9.0",
|
| 3 |
"calibration": null,
|
| 4 |
+
"created_at": "2026-10-06T21:01:32.201103+00:00",
|
| 5 |
"effective_bpw": 6.336767613767534,
|
| 6 |
"files": [
|
| 7 |
{
|
|
|
|
| 161 |
},
|
| 162 |
{
|
| 163 |
"path": "README.md",
|
| 164 |
+
"sha256": "23ec002786a5d6fec5d4ae92dcbff29dc37803a044145794273f22c1663ff213",
|
| 165 |
+
"size_bytes": 8391
|
| 166 |
},
|
| 167 |
{
|
| 168 |
"path": "axquant_plan.json",
|
|
|
|
| 229 |
"sha256": "45fc17c8dd2474af6b493b52483c26c0584b0082d368c480f9fa611e73070040",
|
| 230 |
"size_bytes": 991
|
| 231 |
},
|
| 232 |
+
{
|
| 233 |
+
"path": "runtime_audit.json",
|
| 234 |
+
"sha256": "30e9e9ad6076491d9ce6b74b8f5ce10f8d7c9c6a44311225e32f2e3635e0eb48",
|
| 235 |
+
"size_bytes": 1776
|
| 236 |
+
},
|
| 237 |
{
|
| 238 |
"path": "tokenizer.json",
|
| 239 |
"sha256": "be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506",
|
runtime_audit.json
ADDED
|
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"applied_config_corrections": [],
|
| 3 |
+
"current_config_sha256": "1dde783746ad6bfd1732b8086a932a4dc69f826f4b95368c7a85e333f64d2abc",
|
| 4 |
+
"date": "2026-10-06",
|
| 5 |
+
"header_sha256": {
|
| 6 |
+
"model.safetensors": "7619f985fddeba008519a1ad33a2cdf7722883613881f47be59e41b55df09238"
|
| 7 |
+
},
|
| 8 |
+
"input_bindings": {
|
| 9 |
+
"axquant_manifest.json": "8f018e038cab050689b69e5e6c75aa45fd76f2e2b3c4980b6d04c02dfab923fa",
|
| 10 |
+
"config.json": "1dde783746ad6bfd1732b8086a932a4dc69f826f4b95368c7a85e333f64d2abc",
|
| 11 |
+
"model.safetensors.index.json": "dde816bfa1af02f5fac1adcd1d54f94538d8817abe8cc6423700a32a7fa992cd"
|
| 12 |
+
},
|
| 13 |
+
"issues": [],
|
| 14 |
+
"model_type": "qwen3_vl",
|
| 15 |
+
"mtp_files": [],
|
| 16 |
+
"ngram_action": "none; do not invent n-gram data",
|
| 17 |
+
"ngram_files": [],
|
| 18 |
+
"ngram_quantization": [],
|
| 19 |
+
"ngram_table_metadata": null,
|
| 20 |
+
"ngram_tensor_count": 0,
|
| 21 |
+
"quality_certified": false,
|
| 22 |
+
"quantization": {
|
| 23 |
+
"quantization": {
|
| 24 |
+
"container_mode": "mxfp4",
|
| 25 |
+
"per_module_modes": {
|
| 26 |
+
"affine": 1,
|
| 27 |
+
"mxfp4": 252
|
| 28 |
+
},
|
| 29 |
+
"physical_recipe_issues": []
|
| 30 |
+
},
|
| 31 |
+
"quantization_config": {
|
| 32 |
+
"container_mode": "mxfp4",
|
| 33 |
+
"per_module_modes": {
|
| 34 |
+
"affine": 1,
|
| 35 |
+
"mxfp4": 252
|
| 36 |
+
},
|
| 37 |
+
"physical_recipe_issues": []
|
| 38 |
+
}
|
| 39 |
+
},
|
| 40 |
+
"removed_source_evidence": [],
|
| 41 |
+
"repo_id": "AutomatosX/AX-Qwen3-VL-4B-Instruct-MLX-AXQ-MXFP4",
|
| 42 |
+
"runtime_arch_id": null,
|
| 43 |
+
"runtime_verified": false,
|
| 44 |
+
"schema_version": "axquant.hub-runtime-audit.v1",
|
| 45 |
+
"scope": "Pinned remote config/index/Safetensors header audit; no runtime load or generation claim.",
|
| 46 |
+
"source_revision": "bb64bafcd6440b5f8d8b3d1bc05b587ff012328c",
|
| 47 |
+
"status": "no-audited-peer-export-profile",
|
| 48 |
+
"unchanged_weight_sha256": {
|
| 49 |
+
"model.safetensors": "ec324b13c8cdcc71cd75408e741d78800df41e9232bf59e65764fdab6f854eba"
|
| 50 |
+
}
|
| 51 |
+
}
|