benchmarks.site

Public leaderboard · Agents, tool use & computer use

APEX-Agents

Tests AI agents on long professional tasks like those of bankers, consultants and lawyers.

Leaderboard

Top 3 right now

  1. 1stGemini 4 Carbon (High)84.2% ±4.3%
  2. 2ndGemini 4 Argon (High)82.4% ±4.3%
  3. 3rdFable 5.2 (Max)80.1% ±4.7%

What it measures

480 long-horizon tasks written by investment bankers, consultants and lawyers, done in realistic file and tool environments, scored by Pass@1 against rubrics.

How to read the numbers

Score (%) with margin — higher is better.

Results

5 rows · higher is better (Score)CSV
1Gemini 4 Carbon (High)84.2% ±4.3%
2Gemini 4 Argon (High)82.4% ±4.3%
3Fable 5.2 (Max)80.1% ±4.7%
4Gemini 4 Hydrogen (xHigh)78.0% ±4.8%
5Hy4 (High)77.6% ±4.5%

Note: Only top 5 shown on landing page; full leaderboard behind a link. No update date shown.

Source: Mercor · mercor.com · retrieved 2026-11-17

Markdown · More Agents, tool use & computer use