The AI arena is free today

Open Superagent
AI OrganizationMaintains 12 benchmarks

Scale AI: AI benchmarks & research

Data and evaluation company. Maintains SEAL, MultiChallenge, SWE-bench Pro, MCP-Atlas and PRBench, and co-maintains Humanity's Last Exam and MASK with the Center for AI Safety.

Maintained by Scale AI

12 benchmarks

In one line

Scale AI maintains HLE-Verified (counts toward the LLM Stats Score, 3 models scored), Humanity's Last Exam (counts toward the LLM Stats Score, 116 models scored), Humanity's Last Exam (no tools, text-only) (counts toward the LLM Stats Score, 14 models scored), Humanity's Last Exam (with tools) (counts toward the LLM Stats Score, 6 models scored), Humanity's Last Exam (with tools, text-only) (counts toward the LLM Stats Score, 12 models scored), MASK (counts toward the LLM Stats Score, 1 model scored), MCP Atlas (counts toward the LLM Stats Score, 41 models scored), Multi-Challenge (counts toward the LLM Stats Score, 33 models scored), PRBench-Finance (counts toward the LLM Stats Score, 1 model scored), PRBench-Legal (counts toward the LLM Stats Score, 1 model scored), Seal-0 (counts toward the LLM Stats Score, 7 models scored) and SWE-Bench Pro (counts toward the LLM Stats Score, 63 models scored). It ships no models of its own on LLM Stats.

FAQ

Common questions about Scale AI.

What is Scale AI?

Data and evaluation company. Maintains SEAL, MultiChallenge, SWE-bench Pro, MCP-Atlas and PRBench, and co-maintains Humanity's Last Exam and MASK with the Center for AI Safety. On LLM Stats it maintains 12 benchmarks, led by Humanity's Last Exam, SWE-Bench Pro and MCP Atlas. It ships no models of its own; each score on its benchmarks carries its own provenance label on the model page.

Which Scale AI benchmarks count toward the LLM Stats Score?

Humanity's Last Exam, SWE-Bench Pro, MCP Atlas, Multi-Challenge, Humanity's Last Exam (no tools, text-only), Humanity's Last Exam (with tools, text-only) and 6 more count toward the LLM Stats Score. Only benchmarks admitted to the core set move the main leaderboard; the rest are shown for transparency.

Which models lead Scale AI's benchmarks?

Claude Fable 5.1 leads Humanity's Last Exam at 65.0% (116 models scored), Claude Opus 5.5 leads SWE-Bench Pro at 89.9% (63 models scored) and Muse Spark Open leads MCP Atlas at 89.2% (41 models scored). Each benchmark page lists the full ranking and how the results were produced.

Where are Scale AI's official benchmark results?

Scale AI's official site is https://scale.com. Each benchmark page on LLM Stats links to that benchmark's own paper or leaderboard, and shows the scores next to every other benchmark so models can be compared on one page.