# Generalist AI pretraining corpus (GEN-1)

From the catalogue. Listed by one researcher from one source page.

- id: `generalist-ai-pretraining-corpus-gen-1`
- kind: handheld gripper
- robot or device: wearable and handheld data collection devices ("robot hands"); the GEN-1 post says the pretraining set contains no robot data (class: other)
- how collected: company-run collection network in homes, warehouses and workplaces
- size as stated: thousands of data collection devices shipped across geographies; 500,000 hours claimed, not released
- hours: not stated | hours only claimed: 500,000 | episodes: not stated
- year: 2026
- organisation: Generalist AI
- licence: not stated (class: not stated)
- access: not released (class: not stated)
- link (dataset): https://generalistai.com/blog/gen-1
- note: Proprietary corpus, the largest stated manipulation corpus found; "half a million" entered as 500,000 claimed hours.

Confirm the licence at the link before relying on it.