Instructions to use Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38 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 Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0
# Run inference directly in the terminal:
llama cli -hf Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0
# Run inference directly in the terminal:
llama cli -hf Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

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 Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

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 Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

Use Docker

docker model run hf.co/Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

LM Studio
Jan
Ollama
How to use Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38 with Ollama:
```
ollama run hf.co/Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0
```
Unsloth Desktop

How to use Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

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": "Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38 with Docker Model Runner:
```
docker model run hf.co/Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0
```

Lemonade

How to use Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

Run and chat with the model

lemonade run user.qwen3loop-0.6b-tool-calling-sft-v38-Q8_0

List all available models

lemonade list

Hermes Agent

How to use Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

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 Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38 with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0

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 "Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38:Q8_0" \
  --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"

Qwen3Loop 0.6B — Tool Calling SFT v38 (epoch 2)

Tool-calling specialist (English-only) for agentic harnesses (Pi coding agent via llama.cpp server). Full-parameter SFT from tool_sft_final base, 2 epochs, lr 1.5e-5, 3427 curated samples, loss masks on repeats/schema-violations/ error-follows/false-first-turns (C1/C1b/C1c/C3).

Files

File	What
`model.safetensors` (+`config.json`, `tokenizer.*`, `chat_template.jinja`, `latent_halting_probe.pt`)	Epoch-2 full weights (1.19 GB)
`unrolled_modelo_qwen3loop_tool_sft_v38_ep2_q8_0.gguf`	Unrolled 56-layer Q8_0 for llama.cpp (1.06 GB)

Measured (Q8, temp 0.0, local GPU)

Holdout battery 121: 103/121 (85.1%) — single 30/30, multiple 37/41, python 21/30, abstention 15/20
Probe 25: 13/25 (python 5/5, abstention 0/5)
Multiturn 12: 11/12 · Nav P5 24: 22/24 (temp 0.0 and 0.8, 0 undeclared calls)
Ship bars 5/5 PASS (battery ≥75%, L1 30/30, probe-python ≥3/5, abstention ≥14/20, multiturn ≥10/12)
Live Pi gate (6 locked prompts): 2.5/6 @0.8

Use (llama.cpp router)

llama-server --models-dir models --no-models-autoload --host 127.0.0.1 --port 8080 --jinja -ngl 999 -c 16384

Then load the Q8 in Pi (/llama → /model). Recommended temp 0.8 (greeting refusal and fewer over-triggers vs 0.0).

Limitations (known)

Hard probe-abstention cases (0/5) resist training; multiturn lost 1 (timezone value).
53.5% of called tools appear once (great for generalization, thin per-tool mastery).
2/3427 samples exceed 5000 tokens and never train (silent skip by design).
Training data is English-only; the card declares English operation.

Training data

3427 samples: API/technical 47%, workspace/code 35%, PowerShell 10%, memory 7%, terminal + calibration/abstention/irrelevance batches. Catalog sizes 1–20 (median 9–12); 156 irrelevance-with-catalog refusals; 200 navigation alias copies (30 tool-name variants). Gates: 13 validation rules + eval-overlap (0 HARD) + JSON-schema call validation, all green.

Downloads last month: 189

Safetensors

Model size

0.6B params

Tensor type

F16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Lordnyx/qwen3loop-0.6b-tool-calling-sft-v38

Base model

Qwen/Qwen3-0.6B-Base

Finetuned

Qwen/Qwen3-0.6B

Quantized

(479)

this model