Skip to content
Rise Data Labs
Let's talk

Live model rankings

Measure the frontier.

Where frontier models stand on 17 public benchmarks, from graduate-level science to deep web research, agent-built apps and design. Refreshed daily.

Public benchmarks / Expert perspective

Reasoning

1 benchmarks
Reasoning

GPQA Diamond

Graduate-level science questions that resist search.

Scores updating…

Full leaderboard

Agentic

1 benchmarks
Agentic

Tau-Bench Airline

Tool-using agents handling real customer policies.

Scores updating…

Full leaderboard

App Building

5 benchmarks
Design Arena

Full-Stack Apps

Coding agents building complete apps, judged by people.

Scores updating…

Full leaderboard
Design Arena

Web Apps

Agent-built web applications, head to head.

Scores updating…

Full leaderboard
Design Arena

Mobile Apps

Agent-built mobile apps, voted on by people.

Scores updating…

Full leaderboard
Design Arena

Native Android

Native Android builds from coding agents.

Scores updating…

Full leaderboard
Design Arena

Godot Game Dev

Agents building playable games in Godot.

Scores updating…

Full leaderboard

Design

6 benchmarks
Design Arena

Websites

Head-to-head human votes on generated websites.

Scores updating…

Full leaderboard
Design Arena

UI Components

Which model builds the better interface component.

Scores updating…

Full leaderboard
Design Arena

Game Dev

Browser games generated in one shot, judged by people.

Scores updating…

Full leaderboard
Design Arena

3D

Generated 3D scenes and interactions, head to head.

Scores updating…

Full leaderboard
Design Arena

Data Visualization

Which model turns data into the clearer chart.

Scores updating…

Full leaderboard
Design Arena

SVG

Vector graphics written as code, judged by eye.

Scores updating…

Full leaderboard

Scores from OpenRouter evals, Artificial Analysis and Design Arena, aggregated by OpenRouter.

Beyond public benchmarks

Get a private eval built for your model.

Public leaderboards show where a model ranks. A private eval built by domain experts shows where it breaks, on tasks it has never seen, and what data closes the gap.

  • Unpublished tasks written by vetted US-based subject-matter experts
  • Rubric-based grading with expert review, not just automated scoring
  • Delivered under NDA, with training data to close the gaps it finds
Book a call Browse datasets