## Design Context

### Users
People researching, benchmarking, and operating open-weight language models on Hugging Face. They need a dense view that helps decide which model families and runtime artifacts are worth benchmarking or optimizing.

### Brand Personality
Precise, independent, and work-focused. The interface should feel like a trustworthy research instrument, not a marketing site.

### Aesthetic Direction
Use a light, table-forward dashboard with strong typography, restrained color, and a visible data-flow graphic. The first screen should immediately show that this is a live leaderboard, with summary metrics and filtering controls close to the table.

### Design Principles
- Show useful data first.
- Keep the page credible and scannable.
- Make grouping and filtering obvious.
- Avoid OpenClaw-specific naming or context.
- Treat the bucket-backed HTML page as the product, not a generated report artifact.