Instructions to use albapepper/scoracle-articulator-v4-q6 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use albapepper/scoracle-articulator-v4-q6 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("albapepper/scoracle-articulator-v4-q6")

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 albapepper/scoracle-articulator-v4-q6 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "albapepper/scoracle-articulator-v4-q6"

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": "albapepper/scoracle-articulator-v4-q6"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use albapepper/scoracle-articulator-v4-q6 with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "albapepper/scoracle-articulator-v4-q6"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "albapepper/scoracle-articulator-v4-q6"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "albapepper/scoracle-articulator-v4-q6",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use albapepper/scoracle-articulator-v4-q6 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 "albapepper/scoracle-articulator-v4-q6"

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 albapepper/scoracle-articulator-v4-q6

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use albapepper/scoracle-articulator-v4-q6 with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "albapepper/scoracle-articulator-v4-q6"

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 "albapepper/scoracle-articulator-v4-q6" \
  --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"

Scoracle Articulator v4 Q6

Scoracle Articulator is a compact, on-device sports narrator. It converts structured Scoracle team-data slices into short conversational answers while keeping product names and transport details out of the response.

This is the shipping v4 iter-900 build: a LoRA fine-tune of ibm-granite/granite-4.0-h-1b, fused and quantized to 6-bit MLX weights. The directory is intended for Apple Silicon inference with MLX and for Scoracle's iOS client.

Intended use

The model expects Scoracle's chat template and one of eight compact JSON DATA slices: profile, rating, momentum, results, news, follow-up, mood, or transfer wire. It is not a general-purpose sports knowledge model and should not be asked to supply facts absent from the provided DATA.

Scoracle's runtime wraps generation in a grounding guard. A response containing an unsupported decimal is rejected and retried. Keep that guard enabled in production.

Training and evaluation

Fine-tuning corpus: 1,456 conversational instruction-response pairs over 182 teams, with entity-disjoint validation and holdout sets.
Held-out evaluation: 176 prompts over 22 teams and eight prompt shapes.
Fused 6-bit build: 97.2% grounded numbers; 100% product-name, plumbing, and foreign-team invariants.
Every observed grounding miss in the quantized evaluation was the programmatically detectable decimal-mash class handled by the runtime guard.

Artifact

Format: MLX Safetensors
Quantization: 6-bit affine, group size 64
Approximate download: 1.1GB
Version: v4-iter900-q6-2026-08-29
Base model license: Apache-2.0

The repository includes the tokenizer, chat template, generation configuration, model configuration, weights, and a versioned manifest.json used by the iOS downloader.

Limitations

The model is English-first, narrow-domain, and optimized for short narration of trusted structured inputs. It may generate inaccurate text outside that contract. Scores and claims above describe Scoracle's held-out dataset and runtime settings (temperature=0, maximum 220 generated tokens); they are not general benchmark results.

Downloads last month: 193

Safetensors

Model size

1B params

Tensor type

U32

BF16

MLX

Hardware compatibility

6-bit

Model tree for albapepper/scoracle-articulator-v4-q6

Base model

ibm-granite/granite-4.0-h-1b-base

Finetuned

ibm-granite/granite-4.0-h-1b

Quantized

(45)

this model