CLUE
No estimate
No hardware requirements for this model
The weights for this model have not been published, so it cannot be downloaded or run on your own hardware at any size. It is reachable only through its provider, and no graphics card changes that.
On record
Full specification
Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.
Origin
Who built this model, where, and when it was published.
- Organisation
- Naver Clova,Naver AI Lab
- Organisation type
- Industry,Industry
- Country
- Korea (Republic of)
- Published
- 22 November 2022
- Authors
- Kyuyong Shin, Hanock Kwak, Su Young Kim, Max Nihlen Ramstrom, Jisu Jeong, Jung-Woo Ha, Kyung-Min Kim
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Recommendation
- Task
- Recommender system
Size
How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.
- Parameters
- 160M
- Training data
- tokens
- Epochs
- 8
- Batch size
- 256
160m
Pre-training: We construct a sufficiently large-scale dataset with more than 50B (50741643225) behavior tokens collected over 2 years from search engine and e-commerce platform. Number of behavior tokens 50,741,643,225 (Appendix 1) Downstream: Books: We collect 1,298,489 review logs of 100,000 unique users and 504,572 unique books. Clothing: We collect 928,598 review logs of 100,000 unique users and 314,943 unique clothing, shoes, and jewelry. 50741643225+1298489+928598=50743870312
CLUE is trained with 160M parameters, sequence length (128), and batch size (256).
Training compute
The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.
- Training compute
- 3.5 × 10¹⁸ FLOP
- How it was established
- Reported,Operation counting
using graph reader on Figure 2 b for the upper right point: 0.04 PF-days * 10^15 FLOPS/s* 3600s*24h=3456000000000000000 counting operations: =160000000.00*50743870312*6=4.871411549952 × 10^19 "The computation (PF-days) is calculated as 6 × # of parameters × batch size × # of training steps × sequence length divided by one PF-day = 8.64 × 10^19. We train all models for 100,000 steps."
The training run
What it physically took to train: which chips, how many, for how long, and what that drew from the wall.
- Training hardware
- NVIDIA V100
- Chips used
- 64
- Wall-clock time
- 168 hours (7 days)
- Power draw
- 38.4 kW
We shuffle the dataset at every epoch and train the model for 8 epochs, where the transfer performance begins to plateau. The total training time takes 7 days on 64 V100 GPUs.
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Record confidence
- Likely
Sources
Where this record came from and when it was last checked.
- Reference
- Scaling Law for Recommendation Models: Towards General-purpose User Representations
- Last updated
- 28 November 2025
What the numbers mean
Where it came from
CLUE was published by Naver Clova,Naver AI Lab, in Korea (Republic of), in November 2022. The organisation is categorised as industry,Industry.
It works in Recommendation, and is recorded as doing recommender system.
Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.
Training and provenance
Producing it required around 3.5 × 10¹⁸ FLOP of arithmetic, on NVIDIA V100, which is a statement about the training budget rather than about inference.
Answers
CLUE — common questions
What GPU do I need to run CLUE?
None. CLUE is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.
Is CLUE open source?
The licensing for CLUE was never recorded in our source data. We treat unstated licensing as closed, because an unrecorded licence is not one to rely on.
How many parameters does CLUE have?
CLUE has 160M parameters. 160m. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.
Who created CLUE?
CLUE was published by Naver Clova,Naver AI Lab, based in Korea (Republic of), categorised as industry,Industry.
When was CLUE released?
CLUE was published in November 2022. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.
What is CLUE used for?
CLUE works in Recommendation, and is recorded as handling recommender system. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
How much compute was used to train CLUE?
Around 3.5 × 10¹⁸ FLOP, on NVIDIA V100. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.
The other direction
Looking at it from the other side?
This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.