S4 TPS calculator

Open weights Stanford University 249M parameters October 2021

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 that can run it

818 cards we hold specifications for

Smallest card that fits

Tesla C1080

4 GB · Q8_0 · 148 tok/s

Fastest card

B200

13,607 tok/s · 180 GB

Which GPUs can run S4?

Set the inputs, read the answer

Context length tokens

A longer conversation needs more memory, which can push this model off smaller cards.

Minimum quality

Hides cards that would only fit the model by compressing it below this point.

Search cards

Card memory

Also show cards that cannot run it

818 cards match

Calculating

					Needs	Quantisation	Fit
13,607 tok/s 8,164–21,772 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	1.0 GB	Q8_0	Comfortable
13,607 tok/s 8,164–21,772 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	1.0 GB	Q8_0	Comfortable
10,866 tok/s 6,520–17,385 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	1.0 GB	Q8_0	Comfortable
10,866 tok/s 6,520–17,385 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	1.0 GB	Q8_0	Comfortable
8,690 tok/s 5,214–13,904 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	1.0 GB	Q8_0	Comfortable
8,318 tok/s 4,991–13,308 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.0 GB	Q8_0	Comfortable
8,318 tok/s 4,991–13,308 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.0 GB	Q8_0	Comfortable
7,960 tok/s 4,776–12,737 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	1.0 GB	Q8_0	Comfortable
7,065 tok/s 4,239–11,304 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	1.0 GB	Q8_0	Comfortable
7,065 tok/s 4,239–11,304 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	1.0 GB	Q8_0	Comfortable
7,065 tok/s 4,239–11,304 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	1.0 GB	Q8_0	Comfortable
6,702 tok/s 4,021–10,723 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
5,715 tok/s 3,429–9,144 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
5,715 tok/s 3,429–9,144 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	1.0 GB	Q8_0	Comfortable
5,715 tok/s 3,429–9,144 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
5,715 tok/s 3,429–9,144 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
5,715 tok/s 3,429–9,144 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
4,352 tok/s 2,611–6,963 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	1.0 GB	Q8_0	Comfortable
4,352 tok/s 2,611–6,963 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	1.0 GB	Q8_0	Comfortable
3,626 tok/s 2,176–5,802 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	1.0 GB	Q8_0	Comfortable
3,549 tok/s 2,129–5,678 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	1.0 GB	Q8_0	Comfortable
3,470 tok/s 2,082–5,552 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	1.0 GB	Q8_0	Comfortable
3,470 tok/s 2,082–5,552 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	1.0 GB	Q8_0	Comfortable
3,470 tok/s 2,082–5,552 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	1.0 GB	Q8_0	Comfortable
3,470 tok/s 2,082–5,552 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable

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: Stanford University
Organisation type: Academia
Country: United States of America
Published: 31 October 2021
Authors: Albert Gu, Karan Goel, Christopher Ré

What it does

The problem areas the model was built for. A model can carry several of each.

Domain: Language
Task: Language modeling/generation
Numerical format: FP16

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: 249M
Training data: 103,000,000 tokens
Epochs: 509.02

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: 7.8 × 10¹⁹ FLOP

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
Chips used: 8
Power draw: 6.5 kW

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: Open source

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Why it is tracked: SOTA improvement
Record confidence: Likely
Citations: 3,576
Benchmark data: S4

Sources

Where this record came from and when it was last checked.

Reference: Efficiently Modeling Long Sequences with Structured State Spaces
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run S4

Ranked by estimated tokens per second, newest card first where speeds tie. Because generation is bound by memory bandwidth, this ordering follows bandwidth rather than any gaming benchmark.

The smallest GPUs that still run S4

The cheapest route in, by memory capacity. A tight fit runs the model but leaves nothing spare for a longer conversation.

What the numbers mean

What you need to run it

Minimum card

Tesla C1080

Memory needed

1.0 GB

Fastest

13,607 tok/s

S4 is small enough at 249M parameters that hardware is rarely the obstacle — 818 of the cards we track can run it, including cards several years old.

The least hardware that works is a Tesla C1080. Its 4 GB is enough at Q8_0 compression, giving roughly 148 tokens per second.

At the other end, a B200 generates roughly 13,607 tokens per second on it, on the strength of 8,000 GB/s of memory bandwidth.

Where it came from

S4 was published by Stanford University, in United States of America, in October 2021. academia is the category the publisher falls under.

It works in Language, and is recorded as doing language modeling/generation.

Published weights mean the model runs on your machine rather than someone else's, which is what makes the hardware question below answerable at all.

Understanding the speeds

Half the cards that hold it manage more than 382.1 tokens per second, and 818 exceed reading speed outright.

Being dense, it reads all of itself per token, which is why the ordering by speed below follows the ordering by memory bandwidth so closely.

Its internal architecture is not on file, so memory is approximated from the parameter count and marked accordingly. Expect the real figure to differ, more so at long context.

Training and provenance

The training run consumed about 7.8 × 10¹⁹ FLOP, on NVIDIA A100. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

It was trained on about 103,000,000 tokens of text.

The reason it appears in this catalogue at all is sOTA improvement.

Step by step

How to choose a GPU for S4

The table above has already assessed every card we hold specifications for against this model. Getting to your answer takes six steps.

01

Check what it needs before anything else

The table lists every card that can hold S4 — around 1.0 GB at Q8_0. That figure, not the card's headline performance, is what decides whether it runs.
02

Decide how long your conversations run

Set the context to what you will actually use. The cache grows with the conversation, and it is the usual reason S4 stops fitting a card that seemed fine.
03

Set a quality floor

The quantisation column varies by card, because a bigger card holds a more accurate copy of S4 — Q8_0 on the smallest card that fits. Set a floor to hold the comparison at one level.
04

Compare tokens per second, not specifications

Sort by speed to see how cards rank for S4. It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 13,607 tok/s.
05

Check the fit verdict before buying

A tight fit runs S4 but leaves nothing spare for a longer conversation; comfortable has headroom. If you expect to grow the context, buy for comfortable.
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 S4 is settled.

Answers

S4 — common questions

Can I run S4 on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 1.0 GB and generating roughly 2,534 tokens per second — a comfortable fit.

Can I run S4 on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 1.0 GB and generating roughly 1,552 tokens per second — a comfortable fit.

Can I run S4 on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 1.0 GB and generating roughly 1,922 tokens per second — a comfortable fit.

Can I run S4 on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 1.0 GB and generating roughly 2,279 tokens per second — a comfortable fit.

Is S4 open source?

Its weights are published, so S4 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.

How many parameters does S4 have?

S4 has 249M parameters. 249M (Table 8). 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 S4?

S4 was published by Stanford University, based in United States of America, categorised as academia.

When was S4 released?

S4 was published in October 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.

What is S4 used for?

S4 works in Language, and is recorded as handling language modeling/generation. These are the areas it was designed around; they describe intent rather than a hard boundary.

Where can I download S4?

The weights for S4 are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.

How much compute was used to train S4?

Around 7.8 × 10¹⁹ FLOP, on NVIDIA A100. 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.

Can I run S4 if it does not fit in my GPU?

Partly. Layers that do not fit sit in system memory and run at a fraction of the speed, so a mostly-offloaded S4 is rarely worth using. Every figure here assumes the whole model is on the card.

Would two GPUs run S4 faster?

A second card roughly doubles the memory available but not the generation rate. With 818 cards already able to run S4 alone, the case for pairing is weak.

Why does the quantisation differ between cards for S4?

Each card is shown running the least-compressed copy it can hold, and S4 appears at 1 different compression levels across the cards that fit it. Bigger cards get the more accurate version.

How accurate are these S4 speed estimates?

These are estimates with real error bars. The fastest result here, 8,164–21,772 tok/s on the B200, could reasonably land anywhere in its published range depending on which runtime you use.

What GPU do I need to run S4?

The smallest card in our catalogue that holds S4 is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 1.0 GB, and produces roughly 148 tokens per second. 818 cards in total can run it.

How fast is S4 on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 13,607 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 818 of the cards that can run S4 clear that.

How much VRAM does S4 need?

About 1.0 GB at Q8_0 compression, which is what the smallest card that runs it uses. Less compression needs more: the figures in the memory column above are recalculated for each card, because each one holds the least-compressed version it can.

Source

Original publication

Record last updated 25 May 2026

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.

All AI models