PinchBench Leaderboard

Run the benchmark yourself →

A streamlined, crab-themed benchmarking leaderboard for comparing LLM models as OpenClaw coding agents. Built with Next.js 16, React 19, and Tailwind CSS.

Features

Main Leaderboard (/)

• Clean tabbed interface for Success Rate, Speed, and Cost views
• Crab-themed rankings - Lobster for #1, Crab for #2, Shrimp for #3
• Visual bar chart showing model performance at a glance
• Simplified table with essential metrics only
• Color-coded scores (green/yellow/red) for quick assessment
• Provider color coding with brand colors (Anthropic, OpenAI, Google, etc.)
• Minimal, data-focused design inspired by SkateBench

Submission Detail Page (/submission/[id])

• Circular score gauge showing overall performance percentage
• Category breakdown with scores by task type (Calendar, Coding, Research, etc.)
• Expandable task cards with detailed criterion-by-criterion scoring
• Grading type badges (Automated, LLM Judge, Hybrid)
• Status indicators for success, warnings, and timeouts
• Metadata display including OpenClaw version and submission timestamp

Design Features

• Dark, minimal theme with focus on data clarity
• Crab and lobster emojis throughout for fun, themed experience
• Streamlined layout - removed unnecessary sections
• Tab-based navigation for different metric views
• Simple bar charts for quick visual comparison
• Clean typography with monospace code fonts
• Mobile-responsive design

Tech Stack

• Framework: Next.js 16 (App Router)
• UI Library: React 19.2
• Styling: Tailwind CSS with custom design tokens
• Components: shadcn/ui (Radix UI primitives)
• Icons: Lucide React
• Date Formatting: date-fns
• Type Safety: TypeScript

Project Structure

├── app/
│   ├── page.tsx                    # Main leaderboard page
│   ├── submission/[id]/page.tsx    # Submission detail page
│   ├── layout.tsx                  # Root layout
│   └── globals.css                 # Global styles & theme
├── components/
│   ├── leaderboard-table.tsx       # Main table component
│   ├── task-breakdown.tsx          # Expandable task list
│   ├── score-gauge.tsx             # Circular score visualization
│   └── ui/                         # shadcn/ui components
├── lib/
│   ├── types.ts                    # TypeScript interfaces
│   ├── mock-data.ts                # Sample benchmark data
│   └── utils.ts                    # Utility functions

Getting Started

Prerequisites

• Node.js 18+
• pnpm (recommended)

Installation

1. Install dependencies:

   pnpm install

2. Run the development server:

   pnpm dev

3. Open your browser: Navigate to http://localhost:3000

Build for Production

pnpm build
pnpm start

Deploying to Cloudflare Pages (Recommended)

This project is a Next.js App Router app and should be deployed with Cloudflare Pages, not Workers. The error you saw (Missing entry-point to Worker script or to assets directory) happens when Wrangler is used without a Worker entry-point.

Pages Build Settings

• Framework preset: Next.js
• Build command: pnpm run build
• Build output directory: .next
• Root directory: /

Notes

• Do not use npx wrangler deploy for Pages. Pages builds and deploys automatically from the repo.
• If you already created a Workers project, create a new Pages project and connect this repo.
• Node.js 18+ and pnpm are supported by Pages; no extra config is required for this repo.

Connecting to Real APIs

Currently, the app uses mock data from lib/mock-data.ts. To connect to real benchmark APIs:

1. Create API Route Handlers

Create files in app/api/:

// app/api/leaderboard/route.ts
export async function GET() {
  const response = await fetch("YOUR_API_ENDPOINT/leaderboard");
  const data = await response.json();
  return Response.json(data);
}

// app/api/results/[id]/route.ts
export async function GET(
  request: Request,
  { params }: { params: Promise<{ id: string }> },
) {
  const { id } = await params;
  const response = await fetch(`YOUR_API_ENDPOINT/results/${id}`);
  const data = await response.json();
  return Response.json(data);
}

2. Replace Mock Data

Update components to fetch from API routes instead of importing mock data:

// In app/page.tsx
const response = await fetch("/api/leaderboard");
const entries = await response.json();

3. Add Environment Variables

Create .env.local:

NEXT_PUBLIC_API_URL=https://your-api-domain.com
API_SECRET_KEY=your-secret-key

Customization

Theme Colors

Edit app/globals.css to customize the color scheme:

:root {
  --background: 0 0% 7%;
  --primary: 217 91% 60%;
  --success: 142 71% 45%;
  --warning: 38 92% 50%;
  /* ... */
}

Provider Colors

Edit lib/types.ts to add/modify provider brand colors:

export const PROVIDER_COLORS: Record<string, string> = {
  anthropic: "#d97757",
  openai: "#10a37f",
  google: "#4285f4",
  // Add more providers
};

Task Categories

Add new task categories in lib/types.ts:

export const CATEGORY_ICONS: Record<string, string> = {
  calendar: "📅",
  coding: "💻",
  // Add more categories
};

Key Components

LeaderboardTable

Responsive table with sorting, filtering, and mobile card layout.

Props:

• entries: LeaderboardEntry[] - Array of leaderboard entries

TaskBreakdown

Expandable accordion showing detailed task results with criterion breakdowns.

Props:

• tasks: TaskResult[] - Array of task results

ScoreGauge

Circular progress indicator showing overall score percentage.

Props:

• score: number - Current score
• maxScore: number - Maximum possible score

Data Types

See lib/types.ts for complete TypeScript interfaces:

• LeaderboardEntry - Main leaderboard row data
• TaskResult - Individual task performance
• Submission - Complete submission with all task details

Accessibility

• Semantic HTML elements (<main>, <header>, <table>)
• Keyboard navigation support
• WCAG AA contrast ratios
• Screen reader friendly labels
• Color + icon indicators (not color alone)

Performance

• Server-side rendering with Next.js 16
• Optimized images and assets
• Code splitting and lazy loading
• Turbopack for faster builds

Browser Support

• Chrome/Edge (latest)
• Firefox (latest)
• Safari (latest)
• Mobile browsers (iOS Safari, Chrome Mobile)

License

MIT

Credits

Built with v0 for OpenClaw