Flamingo
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
- DeepMind
- Organisation type
- Industry
- Country
- United Kingdom of Great Britain and Northern Ireland
- Published
- 29 April 2022
- Authors
- Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katie Millican, Malcolm Reynolds, Roman Ring, Eliza Rutherford, Serkan Cabi, Tengda Han, Zhitao Gong, Sina Samangooei, Marianne Monteiro, Jacob Menick, Sebastian Borgeaud, Andrew Brock, Aida Nematzadeh, Sahand Sharifzadeh, Mikolaj Binkowski, Ricardo Barreira, Oriol Vinyals, Andrew Zi…
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Multimodal, Vision, Language, Video
- Task
- Visual question answering, Image captioning
- Approach
- Supervised
- Base model
- Chinchilla
- Numerical format
- BF16
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
- 80B
- Training data
- 458,333,333,333 tokens
"We obtain three models, Flamingo-3B, Flamingo-9B and Flamingo-80B" " The Flamingo-80B model builds on top of the frozen Chinchilla 70B language model [42]. Starting from the very first layer and before every seventh transformer blocks, we add a GATED XATTN-DENSE layer attending to the visual inputs; this accounts for 10B additional learned parameters. For simplicity, we refer to this model as simply Flamingo throughout the paper"
Flamingo was trained on a mixture of web-scraped datasets: 43M pages of text with interleaved images (MultiModal MassiveWeb dataset) 312M image-text pairs (LTIP dataset) 27M video-text pairs (VTP dataset) 1.8B image-alt text pairs (ALIGN dataset) Training dataset size is at least 2.1 billion.
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
- 2.2 × 10²³ FLOP
- How it was established
- Hardware
1536 TPU v4 chips for 15 days. Assuming 40% utilization: C = 1536 TPU * 275*10^12 FLOP/s/TPU * 15 day * 86400 s/day * 0.40 = 2.2*10^23 FLOP "All training and evaluation was performed on TPUv4 instances. The largest model containing 80 billion parameters is trained on QUSV chips for 15 days and sharded across 16 devices." "All trained parameters and optimizer accumulators are stored and updated in float32; all activations and gradients are computed in bfloat16 after downcasting of parameters fr…
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
- Google TPU v4
- Chips used
- 1,536
- Chip-hours
- 552,960
- Wall-clock time
- 360 hours (15 days)
- Power draw
- 1.0 MW
- Compute cost
- $183,423
1536 TPU v4 chips for 15 days
Availability
Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.
- Weights
- Closed — provider access only
- Model access
- Unreleased
- Training code
- Unreleased
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Foundation model
- Yes
- Likely above 10²³ FLOP
- Yes
- Why it is tracked
- SOTA improvement,Discretionary
- Record confidence
- Confident
- Citations
- 5,793
Figure 2 "For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data."
Sources
Where this record came from and when it was last checked.
- Reference
- Flamingo: a Visual Language Model for Few-Shot Learning
- Last updated
- 25 May 2026
What the numbers mean
Where it came from
Flamingo was published by DeepMind, in the country recorded as United Kingdom of Great Britain and Northern Ireland, during April 2022. It comes out of an organisation categorised as industry.
It works in the domain of Multimodal, Vision, Language, Video, and is recorded as performing the task of visual question answering, Image captioning.
Rather than being trained from scratch, it is derived from Chinchilla. That is why it shares the base model's general shape and size.
Its weights were never published, so it can only be reached through its provider. No graphics card changes that.
Training and provenance
The training run consumed about 2.2 × 10²³ FLOP, on hardware recorded as Google TPU v4. That figure measures what producing the model cost, and has no bearing on how fast it answers.
It was trained on a corpus of about 458,333,333,333 tokens of text.
The reason it appears in this catalogue at all: sOTA improvement,Discretionary.
Answers
Flamingo — common questions
Flamingo— who created it?
It was published by DeepMind, based in United Kingdom of Great Britain and Northern Ireland, an organisation categorised as industry.
Flamingo— when was it released?
It was published in April 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.
Flamingo— what is it used for?
It works in the domain of Multimodal, Vision, Language, Video, and is recorded as handling the task of visual question answering, Image captioning. These are the areas it was designed around; they describe intent rather than a hard boundary.
Flamingo— how much compute was used to train it?
Training consumed around 2.2 × 10²³ FLOP, on hardware recorded as Google TPU v4. 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.
Flamingo— what GPU do I need to run it?
None. This 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.
Flamingo— is it open source?
No. Its weights have not been published, so it exists only as a service controlled by its owner.
Flamingo— how many parameters does it have?
It has a parameter count of 80B. "We obtain three models, Flamingo-3B, Flamingo-9B and Flamingo-80B" " The Flamingo-80B model builds on top of the frozen Chinchilla 70B language model [42]. Starting from the very first layer and before every seventh transformer blocks, we add a GATED XATTN-DENSE layer attending to the visual inputs; this accounts for 10B additional learned parameters. For simplicity, we refer to this model as simply Flamingo throughout the paper". 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.
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.