benchmarks.site

Public leaderboard · Agents, tool use & computer use

Remote Labor Index (RLI)

Tests whether AI can complete real paid freelance projects well enough that a client would accept the work.

Leaderboard

Top 3 right now

  1. 1stFable 5.2 (NEW)41.25
  2. 2ndOpus 5.521.25
  3. 3rdGPT 6 Astra20.83

What it measures

Real paid remote freelance projects, scored by the share of projects where the AI's deliverable would be accepted as the finished work.

How to read the numbers

Automation rate: share of projects completed acceptably (%) — higher is better.

Results

23 rows · higher is better (Automation rate (%))CSV
1Fable 5.2 (NEW)41.25
2Opus 5.521.25
3GPT 6 Astra20.83
4Fable 5.117.92
5Gemini Argon 417.08
6Fable 515.83
7Deepseek V4.1 Flash (NEW)9.58
8Opus 4.88.33
9Codex GPT 5.56.25
10Gemini 3.8 Flash5.83
11Gemini 3.7 Flash5.00
12claude-opus-4-6 (CoWork)4.17
13claude-opus-4-5-20251101-thinking3.75
14Manus_1.6 (Max)2.92
15gpt-5.2-2025-12-11 (medium)2.50
15Manus 1.02.50
15Manus 1.52.50
18claude-4-5-Sonnet2.08
18gpt-5.2-2025-12-11 (default)2.08
20gpt-5-2025-08-071.67
21ChatGPT agent1.25
21gemini-3-pro-preview1.25
23gemini-2.5-pro-preview-06-050.83

Note: No update date shown; includes entries tagged NEW. Tied ranks as on page.

Source: Scale AI / Center for AI Safety · labs.scale.com · retrieved 2026-11-17

Markdown · More Agents, tool use & computer use