AI models
Cube-Space AutoEncoder
Training compute
1.1×10¹⁷ FLOP
Published
Apr 27, 2020
Cube-Space AutoEncoder is an AI model developed by MIT-IBM Watson AI Lab (United States), first published in April 2020. It works in the vision and search domain, on tasks such as visual puzzles.
Training it took an estimated 1.1×10¹⁷ FLOP of compute (estimation method: hardware). It was trained on roughly 4.2B datapoints. Training ran on 1 NVIDIA Tesla K80 for about 24 hours.
The reference paper has 56 citations. Epoch AI rates the confidence of this record as confident.
Full record
- Organization
- MIT-IBM Watson AI Lab
- Country of organization
- United States
- Domain
- Vision, Search
- Task
- Visual puzzles
- Training compute
- 1.1×10¹⁷ FLOP
- Compute estimation method
- Hardware
- Dataset size
- 4.2B
- Training hardware
- NVIDIA Tesla K80
- Chips used
- 1
- Training time
- 24 h
- Chip-hours
- 24
- Training power draw
- 337 W
- Citations
- 56
- Epoch confidence
- Confident
SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.
← All ai models