Text Generation
MLX
Safetensors
qwen3_5
apple-silicon
quantized
mixed-precision
axquant
axq
development
qwen3.8
MXFP4
mtp
vision
conversational
4-bit precision
Instructions to use AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP") prompt = "Write a story about Einstein" messages = [{"role": "user", "content": prompt}] prompt = tokenizer.apply_chat_template( messages, add_generation_prompt=True ) text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Pi
How to use AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP 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-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP"
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-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP" } ] } } }Run Pi
# Start Pi in your project directory: pi
- MLX LM
How to use AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Interactive chat REPL mlx_lm.chat --model "AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP"
Run an OpenAI-compatible server
# Install MLX LM uv tool install mlx-lm # Start the server mlx_lm.server --model "AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP" # Calling the OpenAI-compatible server with curl curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP", "messages": [ {"role": "user", "content": "Hello"} ] }' - Hermes Agent
How to use AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP 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-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP"
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-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP 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-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP"
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-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP" \ --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 +64 -0
README.md
CHANGED
|
@@ -19,6 +19,17 @@ tags:
|
|
| 19 |
- vision
|
| 20 |
---
|
| 21 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
# AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP — 4.84 BPW measured main
|
| 23 |
|
| 24 |
An **AXQuant (AXQ)** mixed-precision MLX checkpoint for Apple Silicon, converted directly from
|
|
|
|
| 19 |
- vision
|
| 20 |
---
|
| 21 |
|
| 22 |
+
## Runtime format audit (2026-10-06)
|
| 23 |
+
|
| 24 |
+
No quantization-container correction was needed.
|
| 25 |
+
This family has no n-gram tensors; no n-gram file or declaration was added.
|
| 26 |
+
This remote format audit does not grant an oMLX/MTPLX runtime profile or a successful load/generation claim.
|
| 27 |
+
|
| 28 |
+
See [runtime_audit.json](runtime_audit.json) for pinned config/index/header
|
| 29 |
+
bindings, architecture, physical-format findings, and applied corrections.
|
| 30 |
+
This is development evidence; no quality, MTP exactness, speed, or certification
|
| 31 |
+
claim is added. Historical evidence stays bound to its original revision.
|
| 32 |
+
|
| 33 |
# AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP — 4.84 BPW measured main
|
| 34 |
|
| 35 |
An **AXQuant (AXQ)** mixed-precision MLX checkpoint for Apple Silicon, converted directly from
|
axquant_manifest.json
CHANGED
|
@@ -1,13 +1,13 @@
|
|
| 1 |
{
|
| 2 |
"axquant_version": "1.9.0",
|
| 3 |
"calibration": null,
|
| 4 |
-
"created_at": "2026-10-
|
| 5 |
"effective_bpw": 5.6719817910099914,
|
| 6 |
"files": [
|
| 7 |
{
|
| 8 |
"path": "README.md",
|
| 9 |
-
"sha256": "
|
| 10 |
-
"size_bytes":
|
| 11 |
},
|
| 12 |
{
|
| 13 |
"path": "axquant_mtp_sidecar_manifest.json",
|
|
@@ -99,6 +99,11 @@
|
|
| 99 |
"sha256": "27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516",
|
| 100 |
"size_bytes": 390
|
| 101 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 102 |
{
|
| 103 |
"path": "tokenizer.json",
|
| 104 |
"sha256": "06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523",
|
|
|
|
| 1 |
{
|
| 2 |
"axquant_version": "1.9.0",
|
| 3 |
"calibration": null,
|
| 4 |
+
"created_at": "2026-10-06T21:01:43.468877+00:00",
|
| 5 |
"effective_bpw": 5.6719817910099914,
|
| 6 |
"files": [
|
| 7 |
{
|
| 8 |
"path": "README.md",
|
| 9 |
+
"sha256": "0821757f35bdeeea75401cdd530ce7832da4dec0965db8b0da44d701e8c12545",
|
| 10 |
+
"size_bytes": 9844
|
| 11 |
},
|
| 12 |
{
|
| 13 |
"path": "axquant_mtp_sidecar_manifest.json",
|
|
|
|
| 99 |
"sha256": "27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516",
|
| 100 |
"size_bytes": 390
|
| 101 |
},
|
| 102 |
+
{
|
| 103 |
+
"path": "runtime_audit.json",
|
| 104 |
+
"sha256": "a48762e34b6b2c1c469a9cf612d9cb1bdf9a63b95285c32322fa44b077f74a9c",
|
| 105 |
+
"size_bytes": 2944
|
| 106 |
+
},
|
| 107 |
{
|
| 108 |
"path": "tokenizer.json",
|
| 109 |
"sha256": "06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523",
|
runtime_audit.json
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"applied_config_corrections": [],
|
| 3 |
+
"current_config_sha256": "5fb63c8d952c521b289e1fd412a1e225020301c37cebcb44dfa881e334003619",
|
| 4 |
+
"date": "2026-10-06",
|
| 5 |
+
"header_sha256": {
|
| 6 |
+
"model-00001-of-00003.safetensors": "c31588f384c4e02e86524877e979c460cbd26baff39bbf8f77af0abbc2679ebe",
|
| 7 |
+
"model-00002-of-00003.safetensors": "f37756d1660edff7d27ce4546eb86b6d739a29299c73e35700deb6ebedbd79a0",
|
| 8 |
+
"model-00003-of-00003.safetensors": "c64bfcc96da0bdccbc01bf5d4604e12d70e72a78ec7b3d5372edcc78ed1f4b52",
|
| 9 |
+
"mtp.safetensors": "9538d9f9a614965890da882a8190a619b054f630c2f6858395db40c8f19a56a5",
|
| 10 |
+
"vision.safetensors": "1abf5a82c2d072b89efdad6f24a2221a14fbb896aceb88581b94b6c5e38cf503"
|
| 11 |
+
},
|
| 12 |
+
"input_bindings": {
|
| 13 |
+
"axquant_manifest.json": "6247792a3a68568daefb445ff7d6ae65224ecfb21f03c9545cdd0d73226528d3",
|
| 14 |
+
"axquant_mtp_sidecar_manifest.json": "65b5b43ba1951fbe2be29db9113643ed04a968aef4f36f93b3b3f7819e400504",
|
| 15 |
+
"axquant_omlx_compat.json": "211a3b00bde87ce684d2817c6250738488fa308a9af21e199f9e00b5360d70a5",
|
| 16 |
+
"config.json": "5fb63c8d952c521b289e1fd412a1e225020301c37cebcb44dfa881e334003619",
|
| 17 |
+
"model.safetensors.index.json": "e20dbf6b96058c7b6ccfc1d79745c3f5f4cf8ea4d287ed408da95bdaa4833911",
|
| 18 |
+
"mtplx_runtime.json": "840c099003ecd5e25bcf8cbad13c83ff2335bfb9aebf5d743bfd82b4c0c058ae"
|
| 19 |
+
},
|
| 20 |
+
"issues": [],
|
| 21 |
+
"model_type": "qwen3_5",
|
| 22 |
+
"mtp_files": [
|
| 23 |
+
"mtp.safetensors"
|
| 24 |
+
],
|
| 25 |
+
"ngram_action": "none; do not invent n-gram data",
|
| 26 |
+
"ngram_files": [],
|
| 27 |
+
"ngram_quantization": [],
|
| 28 |
+
"ngram_table_metadata": null,
|
| 29 |
+
"ngram_tensor_count": 0,
|
| 30 |
+
"quality_certified": false,
|
| 31 |
+
"quantization": {
|
| 32 |
+
"quantization": {
|
| 33 |
+
"container_mode": "affine",
|
| 34 |
+
"per_module_modes": {
|
| 35 |
+
"affine": 2,
|
| 36 |
+
"mxfp4": 496
|
| 37 |
+
},
|
| 38 |
+
"physical_recipe_issues": []
|
| 39 |
+
},
|
| 40 |
+
"quantization_config": {
|
| 41 |
+
"container_mode": "affine",
|
| 42 |
+
"per_module_modes": {
|
| 43 |
+
"affine": 2,
|
| 44 |
+
"mxfp4": 496
|
| 45 |
+
},
|
| 46 |
+
"physical_recipe_issues": []
|
| 47 |
+
}
|
| 48 |
+
},
|
| 49 |
+
"removed_source_evidence": [],
|
| 50 |
+
"repo_id": "AutomatosX/AX-Qwen3.8-27B-MLX-AXQ-MXFP4-MTP",
|
| 51 |
+
"runtime_arch_id": "qwen3-next-mtp",
|
| 52 |
+
"runtime_verified": false,
|
| 53 |
+
"schema_version": "axquant.hub-runtime-audit.v1",
|
| 54 |
+
"scope": "Pinned remote config/index/Safetensors header audit; no runtime load or generation claim.",
|
| 55 |
+
"source_revision": "886785fd49428bb33f28198fed53a673572bb7d6",
|
| 56 |
+
"status": "no-audited-peer-export-profile",
|
| 57 |
+
"unchanged_weight_sha256": {
|
| 58 |
+
"model-00001-of-00003.safetensors": "cf4901a75d3c819b69df7002d8134d81982c94d20ba600522add8a2df858033b",
|
| 59 |
+
"model-00002-of-00003.safetensors": "e710ff4cc5964f52e422be8881cd17eacd007831d67a903ce52f3d7452dc29e8",
|
| 60 |
+
"model-00003-of-00003.safetensors": "5cd45370f41b0d6bf5b8f10def7b0be1cbaf76cdc392320c5778726763a639f5",
|
| 61 |
+
"mtp.safetensors": "914bef4ec6fcfde83925f9c138559253db030d9138b0a375623d6fdc06ebe898",
|
| 62 |
+
"vision.safetensors": "d0d927c489250588557d5761dcd47ca267ef64c95f7e7d58dd6a0cccf54ac68f"
|
| 63 |
+
}
|
| 64 |
+
}
|