Instructions to use AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX 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("AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX")

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

How to use AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX"

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": "AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX 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 "AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX"

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 AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX"

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 "AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX" \
  --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"

NOESIS / AMAImedia

Last updated: 2026-08-29

Released as part of the NOESIS Professional Multilingual Dubbing Automation Platform (framework: DHCF-FNO — Deterministic Hybrid Control Framework for Frozen Neural Operators).

Founder: Ilia Bolotnikov
Organization: AMAImedia.com
X (Twitter): @AMAImediacom
LinkedIn: Ilia Bolotnikov
Telegram: @djbionicl
NOESIS version: v16.1
Release date: 2026-08-26

NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX

Original NOESIS MLX model for lightweight inspection, validation, routing checks, and agent-quality review. This repository contains the local MLX safetensors conversion of the Qwopus3.5 v3 Inspector specialist. It is intended for Apple Silicon inference through mlx-lm; the files may be stored on Windows, but MLX execution requires a compatible macOS/Apple-Silicon environment.

This is the MLX counterpart of the published NOESIS-Qwopus3.5-0.8B-v3-Inspector-Q8_0 model card. The Q8_0 repository is the llama.cpp-compatible artifact; this repository preserves the original NOESIS model in MLX format for Apple Silicon.

Released as part of the NOESIS Professional Multilingual Dubbing Automation Platform (DHCF-FNO — Deterministic Hybrid Control Framework for Frozen Neural Operators).

Property	Value
Model role	Inspector specialist for lightweight validation, review, and agent-quality checks
Architecture	`Qwen3_5ForCausalLM`, `qwen3_5`
Parameters	Approximately 0.8B
Layers / hidden size	24 / 1024
Context limit in config	1,048,576 positions
Stored dtype	BF16 configuration; MLX safetensors
Weight file	`model.safetensors`, approximately 1.50 GB
Tokenizer	Qwen3.5 tokenizer with `chat_template.jinja`
Runtime	`mlx-lm` on Apple Silicon
NOESIS version	v16.1
Release date	2026-08-26

Language support

The Qwen3.5 family advertises expanded coverage of 201 languages and dialects. The upstream Qwen3 language list below is the transparent documented baseline used for this derivative; the published Qwen3.5 count is broader than the enumerated list.

English, French, Portuguese, German, Romanian, Swedish, Danish, Bulgarian, Russian, Czech, Greek, Ukrainian, Spanish, Dutch, Slovak, Croatian, Polish, Lithuanian, Norwegian Bokmål, Norwegian Nynorsk, Persian, Slovenian, Gujarati, Latvian, Italian, Occitan, Nepali, Marathi, Belarusian, Serbian, Luxembourgish, Venetian, Assamese, Welsh, Silesian, Asturian, Chhattisgarhi, Awadhi, Maithili, Bhojpuri, Sindhi, Irish, Faroese, Hindi, Punjabi, Bengali, Oriya, Tajik, Eastern Yiddish, Lombard, Ligurian, Sicilian, Friulian, Sardinian, Galician, Catalan, Icelandic, Tosk Albanian, Limburgish, Dari, Afrikaans, Macedonian, Sinhala, Urdu, Magahi, Bosnian, Armenian; Chinese (Simplified Chinese, Traditional Chinese, Cantonese), Burmese; Arabic (Standard, Najdi, Levantine, Egyptian, Moroccan, Mesopotamian, Ta’izzi-Adeni, Tunisian), Hebrew, Maltese; Indonesian, Malay, Tagalog, Cebuano, Javanese, Sundanese, Minangkabau, Balinese, Banjar, Pangasinan, Iloko, Waray (Philippines); Tamil, Telugu, Kannada, Malayalam; Turkish, North Azerbaijani, Northern Uzbek, Kazakh, Bashkir, Tatar; Thai, Lao; Finnish, Estonian, Hungarian; Vietnamese, Khmer; Japanese, Korean, Georgian, Basque, Haitian, Papiamento, Kabuverdianu, Tok Pisin, Swahili.

The model card intentionally avoids claiming identical quality for every language. For production dubbing, language quality must be checked by the NOESIS multilingual ASR, speaker, emotion, and timing gates.

Repository contents

NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX/
├── README.md
├── model.safetensors              # MLX model weights, ~1.50 GB
├── model.safetensors.index.json   # weight metadata
├── config.json                    # Qwen3.5 architecture configuration
├── generation_config.json
├── tokenizer.json
├── tokenizer_config.json
└── chat_template.jinja            # chat formatting used by the tokenizer

How to run with MLX-LM

Important platform note

mlx-lm is designed for Apple Silicon and uses the MLX backend. The B:\Downloads\... path is the Windows storage location of this repository; it is not a native Windows execution path for MLX. Copy or sync the repository to an Apple Silicon Mac, then use a macOS path such as /Users/ilia/Models/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX.

Install

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -U pip mlx-lm

Single prompt generation

python -m mlx_lm.generate \
  --model /Users/ilia/Models/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX \
  --prompt "Inspect this short transcript for contradictions and return a concise verdict." \
  --max-tokens 256 \
  --temp 0.2

The --model value may be a local directory or a Hugging Face repository ID. For this repository, use the local directory when testing the exact files that were uploaded. If mlx-lm requests permission for tokenizer code, review the prompt and add --trust-remote-code only when required by the installed runtime.

Chat mode

python -m mlx_lm.chat \
  --model /Users/ilia/Models/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX

The tokenizer's chat_template.jinja is used by mlx-lm when chat messages are formatted. The Inspector is intended for short, deterministic review tasks, so prefer low temperature and bounded max_tokens values.

OpenAI-compatible local server

python -m mlx_lm.server \
  --model /Users/ilia/Models/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX \
  --host 127.0.0.1 \
  --port 8080

In another terminal:

curl http://127.0.0.1:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [
      {"role": "system", "content": "You are the NOESIS Inspector. Return concise, evidence-based verdicts."},
      {"role": "user", "content": "Check this subtitle line for an obvious timing or translation risk."}
    ],
    "temperature": 0.2,
    "max_tokens": 256
  }'

The MLX-LM server implements a basic OpenAI-like API on localhost. It is suitable for local development and pipeline experiments, not for direct public exposure.

Python API

from mlx_lm import load, generate

model_dir = "/Users/ilia/Models/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX"
model, tokenizer = load(model_dir)

messages = [
    {"role": "system", "content": "You are the NOESIS Inspector. Be concise and evidence-based."},
    {"role": "user", "content": "Review this dubbing decision and return PASS, WARN, or FAIL with one reason."},
]
prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
answer = generate(
    model,
    tokenizer,
    prompt=prompt,
    max_tokens=256,
    temp=0.2,
    verbose=True,
)
print(answer)

NOESIS role and routing

Inspector is a small specialist, not a replacement for the 9B Supervisor or DubCompress models. Use it for fast pre-checks, lightweight validation, and agent-quality inspection. Escalate complex reasoning, long-context supervision, and final dubbing decisions to the appropriate 9B NOESIS model.

Task	Recommended route
Fast validation, sanity check, compact verdict	Inspector 0.8B MLX
Long-context supervision and multi-agent review	Supervisor 9B MLX
Translation compression and subtitle-slot fitting	DubCompress 9B MLX
External/public deployment	Not recommended; keep the MLX server on localhost

Runtime notes

Item	Guidance
Memory	The weight file is approximately 1.50 GB; leave additional unified memory for tokenizer, KV cache, and the operating system.
Context	The configuration declares a large maximum context, but practical context should be limited by available unified memory and task latency.
Sampling	Use low temperature for inspection and validation. Always cap `max_tokens`.
Chat format	Use `chat_template.jinja` through `tokenizer.apply_chat_template`.
Quality	A successful MLX load is a technical smoke test, not a production-quality gate.

NOESIS sealed rules

Original model preservation: this MLX repository is an original NOESIS model artifact, not an external repack.
R-INSPECTOR-FAST-REVIEW: use Inspector for lightweight checks and escalate difficult judgments to Supervisor.
R-MLX-LOCALHOST: the MLX-LM server must remain local unless a separately secured gateway is provided.
R-CHAT-TEMPLATE-PRESERVE: preserve the shipped chat_template.jinja and tokenizer files together with the weights.
R-APACHE-CLEAN: preserve the applicable upstream and NOESIS license notices.

References

License

See the repository license and upstream notices. The model is released as part of the NOESIS / AMAImedia original-trained model family. Redistribution must preserve the included metadata and attribution.

Downloads last month: -

Safetensors

Model size

0.8B params

Tensor type

BF16

MLX

Hardware compatibility

Quantized

Collection including AMAImedia/NOESIS-Qwopus3.5-0.8B-v3-Inspector-MLX

NOESIS-Original-Trained-Models

Collection

24 items • Updated 31 minutes ago