GCB Runner

Run the Great Commission Benchmark locally on any AI model

Test Any Model, Anywhere

The GCB Runner is the official command-line tool for running the Great Commission Benchmark on your own infrastructure. Perfect for testing local models, fine-tuned models, or models not available through the platform.

OpenRouterOpenAIAnthropicLM StudioOllama

Video coming soon

Quick Install for macOS
Recommended
One command installs everything — no Python required
curl -fsSL https://greatcommissionbenchmark.ai/install.sh | bash

This installer will:

  • Detect your Mac type (Apple Silicon or Intel)
  • Download and verify the latest GCB Runner
  • Handle macOS security settings automatically
  • Install to your PATH for easy access

Works on Apple Silicon (M1/M2/M3/M4) Macs.

Manual Download
Download the standalone executable directly
Download for macOS (Apple Silicon)
Manual installation steps

After downloading:

  1. Open Terminal and navigate to your Downloads folder:cd ~/Downloads
  2. Remove macOS quarantine and make executable:xattr -d com.apple.quarantine gcb-runner-macos-arm64 && chmod +x gcb-runner-macos-arm64
  3. Move to your PATH (optional):sudo mv gcb-runner-macos-arm64 /usr/local/bin/gcb-runner
  4. Run the tool:gcb-runner
Auto-updates included — gcb-runner will notify you when updates are available

Verify downloads using SHA256 hashes from manifest.json

Developer Installation
Install from source for development or customization

For developers who want to run from source, contribute, or customize the runner:

# Clone the repository
git clone https://github.com/GospelAmbition/gcb-runner.gitcd gcb-runner
# Install with pip (editable mode)
pip install -e ".[dev]"
# Run the tool
gcb-runner
View on GitHub
Requirements: Python 3.10 or higher

Features

Local Models

Test models running on LM Studio, Ollama, or any OpenAI-compatible endpoint. Perfect for fine-tuned models or private deployments.

Cloud APIs

Connect to OpenRouter, OpenAI, or Anthropic APIs directly. Access 100+ models with your own API keys and usage limits.

Local Dashboard

View results in a local web dashboard. Generate HTML reports, compare runs, and analyze performance before submitting.

Submit to Leaderboard

Upload verified results to the platform leaderboard. Submissions are reviewed by moderators to ensure benchmark integrity.

Quick Start
# Launch interactive menu
gcb-runner
# Or use commands directly
gcb-runner config # Set up API keysgcb-runner test --model gpt-4o --backend openrouter # Run benchmarkgcb-runner view # Open dashboardgcb-runner export # Export results