Instructions to use SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit 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("SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit")

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 SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit"

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": "SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit 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 "SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit"

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 SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit"

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 "SoftBacon-Software/Laguna-XS-2.1-sbQ-4bit" \
  --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"

Laguna-XS-2.1-sbQ-4bit

File size: 896 Bytes

71fe2ff

# PROVENANCE — Laguna-XS-2.1-mlx-oq4e-ours

**This is OUR quantization of an OFFICIAL poolside release. It is not a
poolside artifact and must not be described as one.**

| field | value |
|---|---|
| upstream model | `poolside/Laguna-XS-2.1` (OpenMDW-1.1) |
| upstream revision | `local-copy (see its PROVENANCE.md)` |
| local source | `/Users/grb/.omlx/models/Laguna-XS-2.1-bf16` |
| produced by | `jarvis/tools/oq_factory.py --arm oq4e` |
| quantization | oQ4 (oQ4e imatrix-weighted) |
| imatrix corpus | built-in oqe corpus |
| omlx | `782ef068c088` on `m5max/oq-calibration` |
| mlx / mlx_lm | 0.32.0 / 0.31.3 |
| output size | 19.5 GB |
| built | 2026-08-24T01:16:32+00:00 |
| wall time | 6.3 min |

Machine-readable copy: `oq_build.json`. Do not promote this build on size
or on this document — the held-out NLL ladder
(`training/glm-distill/eval/run_quant_ladder.sh`) is the arbiter.