Atlas · C02 · Body

Fine manipulation & dexterity

Precise, adaptive use of hands and fingers on varied, deformable or unfamiliar objects.

AI Reliable score

2 / 10

Beginner level · peak 4 · trend ↑ · grade B

Why this score

Verified work is narrow and structured (BMW pilot: sheet-metal placement within 5 mm). General household tasks still fail most of the time (88% failure rate cited for 2025-26 research); complex home tasks rely on remote human operators. Heavy investment, slow verified progress.

On the human scale

2Beginner. Grasps and places rigid objects slowly ← AI today

4Average adult. Average adult: buttons, cooking, everyday tool use

6Competent pro. Trade professional: assembles, wires, sews, handles varied parts at work pace

8Expert. Expert: surgeon / watchmaker precision with adaptation

10Best humans. World-class micro-surgeons, virtuoso musicians

Future warning

At its best, AI already reaches 4 on this capability. Peak results are never counted in the index; they show where the water may rise next.

Evidence

Humanoid robot deployment reviews 2026C

E03 · 2026

Measures: Verified real-world humanoid work
AI result: Figure 02 at BMW: ~1,250 h, 90,000+ parts placed within 5 mm. 88% household-task failure rate cited for 2025-26 research. 1X NEO uses remote operators for complex tasks. Batteries 3-5 h; no humanoid works a full untethered shift.
Humans: Human workers (no head-to-head figures published)
On our scale: Structured, supervised, narrow tasks only: beginner level for general dexterity.

robobrief.tech ↗ dev.to ↗

Visuo-tactile manipulation benchmarks (Bench2Dex, RoboMNI)C

E04 · 2026-09

Measures: Touch-based perception and dexterous manipulation
AI result: New research benchmarks; no production evidence of human-level touch.
Humans: Not established
On our scale: Research stage: Reliable 2.

huggingface.co ↗ www.therobotreport.com ↗

History

First reading: Atlas v0.1, 5 Nov 2026. Every later change will be listed here with its reason.

Disagree? Send a correction