Instructions to use aitups/Qwen3.5-4B-saor with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use aitups/Qwen3.5-4B-saor with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf aitups/Qwen3.5-4B-saor
# Run inference directly in the terminal:
llama cli -hf aitups/Qwen3.5-4B-saor

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf aitups/Qwen3.5-4B-saor
# Run inference directly in the terminal:
llama cli -hf aitups/Qwen3.5-4B-saor

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf aitups/Qwen3.5-4B-saor
# Run inference directly in the terminal:
./llama-cli -hf aitups/Qwen3.5-4B-saor

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf aitups/Qwen3.5-4B-saor
# Run inference directly in the terminal:
./build/bin/llama-cli -hf aitups/Qwen3.5-4B-saor

Use Docker

docker model run hf.co/aitups/Qwen3.5-4B-saor

LM Studio
Jan
Ollama
How to use aitups/Qwen3.5-4B-saor with Ollama:
```
ollama run hf.co/aitups/Qwen3.5-4B-saor
```
Unsloth Desktop

How to use aitups/Qwen3.5-4B-saor with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf aitups/Qwen3.5-4B-saor

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "aitups/Qwen3.5-4B-saor"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use aitups/Qwen3.5-4B-saor with Docker Model Runner:
```
docker model run hf.co/aitups/Qwen3.5-4B-saor
```

Lemonade

How to use aitups/Qwen3.5-4B-saor with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull aitups/Qwen3.5-4B-saor

Run and chat with the model

lemonade run user.Qwen3.5-4B-saor-{{QUANT_TAG}}

List all available models

lemonade list

Hermes Agent

How to use aitups/Qwen3.5-4B-saor with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf aitups/Qwen3.5-4B-saor

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 aitups/Qwen3.5-4B-saor

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use aitups/Qwen3.5-4B-saor with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf aitups/Qwen3.5-4B-saor

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 "aitups/Qwen3.5-4B-saor" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Qwen3.5-4B-saor

Modelo Qwen3.5-4B con los bloques FFN podados por magnitud (conexiones de mayor |w|) según un perfil de densidad por capa optimizado con CMA-ES (genoma CPPN como generador indirecto). Pesos activos re-empaquetados en Q4_K; adyacencia en formato GGUF disperso D16 (ffn_dag_adjacency + ffn_dag_weights).

Arquitectura base: transformer denso (33 bloques FFN [2560 → 9216]).

Métrica	Valor
Divergencia KL (vs base, n_pos=4)	0.0638
Compresión D_arch (gate / modelo)	0.0965 / 0.0333
Capas esparsas	33
Método de poda	Magnitud (top-\|w\|) + perfil CPPN

El perfil concentra la esparsidad en las capas intermedias (pico ~0.16) y deja las capas iniciales/finales casi densas. Para referencia: la poda uniforme a sp 0.10 da KL 0.108.

Ejecutar con Hayai

# 1. Compilar Hayai
cd hayai && cargo build --release

# 2. Dry-run del plan
cargo run --release -p hayai-cli -- plan --model Qwen3.5-4B-saor.gguf

# 3. Generar texto
cargo run --release -p hayai-cli -- generate \
  --model Qwen3.5-4B-saor.gguf \
  --prompt "The capital of France is" \
  --max-tokens 64 \
  --device auto

Evaluar la divergencia KL

cargo run --release --example kl_eval -- \
  --orig Qwen3.5-4B-Q4_K_M.gguf \
  --sparse Qwen3.5-4B-saor.gguf \
  --prompts calib128.txt --n-positions 4 --device auto

Archivos

Qwen3.5-4B-saor.gguf — el modelo disperso (~2.8 GB, Q4_K).
sparsities.txt — perfil de esparsidad por capa (33 floats).

Reproducibilidad

Perfil obtenido con el loop evolutivo de SAOR (via_b_evolve --batch-eval --magnitude, 12 generaciones): CMA-ES sobre un genoma CPPN que decodifica el perfil de densidad; adyacencia = top-|w| a esa densidad. Genoma (466 f32) e histórico disponibles bajo demanda.

Atribución: modelo base Qwen/Qwen3.5-4B — consulte la licencia del modelo base.