RWKV-4 14B
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
- RWKV Foundation
- Organisation type
- Research collective
- Country
- Multinational
- Published
- 22 May 2023
- Authors
- Bo Peng, Eric Alcaide, Quentin Anthony, Alon Albalak, Samuel Arcadinho, Huanqi Cao, Xin Cheng, Michael Chung, Matteo Grella, Kranthi Kiran GV, Xuzheng He, Haowen Hou, Przemyslaw Kazienko, Jan Kocon, Jiaming Kong, Bartlomiej Koptyra, Hayden Lau, Krishna Sri Ipsit Mantri, Ferdinand Mom, Atsushi Saito, Xiangru Tang, Bolun Wang, Johan S. Wind, Stansilaw Wozniak, Ruichong Zhang, Zhenyuan Zhang, Qihang …
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Language modeling
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
- 14B
- Training data
- 330,000,000,000 tokens
- Batch size
- 262,144
14b
262144 (or 131072?) "To train the models mentioned, we... switch batch size dynamically between 128 or 256 sequences, each of 1024 tokens"
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
- 2.8 × 10²² FLOP
- How it was established
- Operation counting
from HuggingFace page: https://huggingface.co/BlinkDL/rwkv-4-pile-14b trained for 330B tokens 14 billion * 330 billion * 6 = 2.78e22 paper notes that a forward pass is almost exactly 2x parameters (within 2%): "Alternative approximations for FLOPs include doubling the parameters which yields similar results within 2% for 14B and a 30% discrepancy for 169M variant." and that 6*params*tokens is a good approximation because it's not a transformer: "FLOPs is for a forward pass for one token. It wa…
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 A100 SXM4 80 GB
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Record confidence
- Confident
- Citations
- 1,042
Sources
Where this record came from and when it was last checked.
- Reference
- RWKV: Reinventing RNNs for the Transformer Era
- Last updated
- 25 May 2026
What the numbers mean
Where it came from
RWKV-4 14B was published by RWKV Foundation, in Multinational, in May 2023. research collective is the category the publisher falls under.
It works in Language, and is recorded as doing language modeling.
Its weights were never published, so it can only be reached through its provider. No graphics card changes that.
Training and provenance
Training it took roughly 2.8 × 10²² FLOP of computation, on NVIDIA A100 SXM4 80 GB — a measure of what producing the model cost, not of how fast it answers.
The training set ran to roughly 330,000,000,000 tokens.
Answers
RWKV-4 14B — common questions
How much compute was used to train RWKV-4 14B?
Around 2.8 × 10²² FLOP, on NVIDIA A100 SXM4 80 GB. 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.
What GPU do I need to run RWKV-4 14B?
None. RWKV-4 14B 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 RWKV-4 14B open source?
The licensing for RWKV-4 14B 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 RWKV-4 14B have?
RWKV-4 14B has 14B parameters. 14b. 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 RWKV-4 14B?
RWKV-4 14B was published by RWKV Foundation, based in Multinational, categorised as research collective.
When was RWKV-4 14B released?
RWKV-4 14B was published in May 2023. 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 RWKV-4 14B used for?
RWKV-4 14B works in Language, and is recorded as handling language modeling. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.
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.