AI models
DensePhrases
Training compute
2.1×10¹⁸ FLOP
Published
Dec 23, 2020
DensePhrases is an AI model developed by Korea University and Princeton University (South Korea and United States), first published in December 2020. It works in the language domain, on tasks such as question answering.
Training it took an estimated 2.1×10¹⁸ FLOP of compute (estimation method: hardware). It was trained on roughly 58M datapoints. Training ran on 8 NVIDIA TITAN Xp for about 20 hours.
Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 128 citations. Epoch AI rates the confidence of this record as speculative.
Full record
- Organization
- Korea University, Princeton University
- Country of organization
- South Korea, United States
- Domain
- Language
- Task
- Question answering
- Training compute
- 2.1×10¹⁸ FLOP
- Compute estimation method
- Hardware
- Dataset size
- 58M
- Training hardware
- NVIDIA TITAN Xp
- Chips used
- 8
- Training time
- 20 h
- Chip-hours
- 160
- Training power draw
- 4.1 kW
- Model accessibility
- Open weights (unrestricted)
- Open weights
- Yes
- Citations
- 128
- Epoch confidence
- Speculative
More from Korea University,Princeton University
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