Switch TPS calculator
Each card below is assessed against this model at the context length and minimum quality you choose. Speed is an estimate for a single request, calculated from the card's memory bandwidth and the size of the model once compressed.
Calculated for this model
818 cards we hold specifications for
Which GPUs can run Switch?
Set the inputs, read the answer
A longer conversation needs more memory, which can push this model off smaller cards.
Hides cards that would only fit the model by compressing it below this point.
0 cards match
Calculating| Needs | Quantisation | Fit | |||||
|---|---|---|---|---|---|---|---|
|
No card in our catalogue can run this model with these settings. |
|||||||
Speeds are estimates for a single request — one conversation at a time — calculated from memory bandwidth, model size and quantisation. Real throughput varies with the inference runtime and its version. Figures published by hardware vendors measure many simultaneous requests and are much higher.
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
- Organisation type
- Industry
- Country
- United States of America
- Published
- 11 January 2021
- Authors
- William Fedus, Barret Zoph, Noam Shazeer
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Text autocompletion
- Approach
- Self-supervised learning
- 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
- 1.6T
- Training data
- 86,400,000,000 tokens
"Combining expert, model and data parallelism, we design two large Switch Transformer models, one with 395 billion and 1.6 trillion parameters" Table 9 gives more precise count of 1571B parameters
"In our protocol we pre-train with 2^20 (1,048,576) tokens per batch for 550k steps amounting to 576B total tokens." 1 token ~ 0.75 words
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
- 8.2 × 10²² FLOP
- How it was established
- Third-party estimation
Table 4 https://arxiv.org/ftp/arxiv/papers/2104/2104.10350.pdf
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 v3
- Chips used
- 1,024
- Chip-hours
- 663,552
- Wall-clock time
- 648 hours (27 days)
- Hardware utilisation
- HFU 28.0%
- Power draw
- 935.4 kW
- Compute cost
- $145,101
see table 4 in https://arxiv.org/ftp/arxiv/papers/2104/2104.10350.pdf
Table 4 in https://arxiv.org/pdf/2104.10350 gives measured performance of 34.4 TFLOP/s, vs. peak achievable FLOP/s of 123 TFLOP/s on the TPUv3 being used. HFU = 34.4/123 = 0.27967
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
- Open — downloadable
- Model access
- Open weights (unrestricted)
- Training code
- Unreleased
Apache 2 for weights: https://huggingface.co/google/switch-c-2048 paper links to this repo but not clear that the training hyperparams for Switch are here: https://github.com/google-research/t5x
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Frontier model
- Yes
- Why it is tracked
- Highly cited,SOTA improvement
- Record confidence
- Confident
- Citations
- 3,888
" On ANLI (Nie et al., 2019), Switch XXL improves over the prior state-of-the-art to get a 65.7 accuracy versus the prior best of 49.4 (Yang et al., 2020)... Finally, we also conduct an early examination of the model’s knowledge with three closed-book knowledge-based tasks: Natural Questions, WebQuestions and TriviaQA, without additional pre-training using Salient Span Masking (Guu et al., 2020). In all three cases, we observe improvements over the prior stateof-the-art T5-XXL model (without SSM…
Sources
Where this record came from and when it was last checked.
- Reference
- Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity
- Last updated
- 25 May 2026
What the numbers mean
What you need to run it
Switch reaches a parameter count of 1.6T. That is beyond what any single graphics card holds. Running it means either splitting it across several cards or renting hardware built for the job, and every card able to hold it alone is a datacentre part. The number that can: 0.
Background
Switch was published by Google, in the country recorded as United States of America, during January 2021. The publishing organisation is categorised as industry.
It works in the domain of Language, and is recorded as performing the task of text autocompletion.
The weights are published, so it can be downloaded and run on your own hardware indefinitely, offline, with no account attached.
Training and provenance
The training run consumed about 8.2 × 10²² FLOP, on hardware recorded as Google TPU v3. That figure measures what producing the model cost, and has no bearing on how fast it answers.
Training consumed a corpus of around 86,400,000,000 tokens of text.
It is tracked in the underlying dataset for one reason in particular: highly cited,SOTA improvement.
Step by step
How to choose a GPU for Switch
The table above has already assessed every card we hold specifications for against this model. Getting to your answer takes six steps.
-
01
Read the memory figure first
Start from what it actually needs, which is the requirement of Switch. That figure, not the headline performance of a card, is what decides whether it runs.
-
02
Decide how long your conversations run
The conversation occupies memory too, and grows as it goes. Set the slider to the length you expect, because at long context a card that handles short questions easily can be dropped by Switch.
-
03
Decide how much compression you will accept
Compression is what makes a model fit smaller cards, at some cost in accuracy. Setting a minimum quality drops the cards that only manage it by squeezing further than you would want, and holds the comparison at one level.
-
04
Sort by speed
Sort by speed to see how cards rank for Switch. It will not match a gaming ordering, because generation is bound by memory bandwidth.
-
05
Read the fit column last
The fit column separates cards that just manage it from those with room to spare, in the case of Switch. Comfortable means you can grow the context later. That difference matters more than a few tokens per second, so buy for comfortable if you expect to.
-
06
See what else that card runs
Following a card through to its own page shows every other model it can hold, which is the question that follows once you have settled on Switch.
Answers
Switch — common questions
Switch— what is it used for?
It works in the domain of Language, and is recorded as handling the task of text autocompletion. These are the areas it was designed around; they describe intent rather than a hard boundary.
Switch— where can I download it?
The weights are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.
Switch— how much compute was used to train it?
Training consumed around 8.2 × 10²² FLOP, on hardware recorded as Google TPU v3. 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.
Switch— can I run it if it does not fit in my GPU?
It can be split between the card and system memory, but it generates painfully slowly that way. The nearest miss we calculate falls short by 691.9 GB. Every figure here assumes the whole model is resident on the card.
Switch— would two GPUs run it faster?
Two cards buy memory rather than speed, which matters only if one card cannot hold it. The number that can: 0. So a second card is rarely the answer here.
Switch— why does the quantisation differ between cards?
A larger card holds a more accurate copy. The number of compression levels used across the cards that run it: 1. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.
Switch— how accurate are these speed estimates?
These are estimates with real error bars, and any of them could reasonably land anywhere in its published range depending on which runtime you use. The fastest result here: the range beneath each figure. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.
Switch— is it open source?
Its weights are published, so it can be downloaded and run on your own hardware. Note that open weights is not the same as open source in the full sense — it says nothing about the training data, the training code, or the commercial terms attached.
Switch— how many parameters does it have?
It has a parameter count of 1.6T. "Combining expert, model and data parallelism, we design two large Switch Transformer models, one with 395 billion and 1.6 trillion parameters" Table 9 gives more precise count of 1571B parameters. 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.
Switch— who created it?
It was published by Google, based in United States of America, an organisation categorised as industry.
Switch— when was it released?
It was published in January 2021. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.
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.