GPT-2 Medium (FlashAttention) TPS calculator

Open weights Stanford University,University at Buffalo 355M parameters May 2022

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 · 104 tok/s

Fastest card

B200

9,544 tok/s · 180 GB

Which GPUs can run GPT-2 Medium (FlashAttention)?

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
9,544 tok/s 5,727–15,271 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	1.1 GB	Q8_0	Comfortable
9,544 tok/s 5,727–15,271 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	1.1 GB	Q8_0	Comfortable
7,621 tok/s 4,573–12,194 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	1.1 GB	Q8_0	Comfortable
7,621 tok/s 4,573–12,194 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	1.1 GB	Q8_0	Comfortable
6,095 tok/s 3,657–9,752 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	1.1 GB	Q8_0	Comfortable
5,834 tok/s 3,500–9,334 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.1 GB	Q8_0	Comfortable
5,834 tok/s 3,500–9,334 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.1 GB	Q8_0	Comfortable
5,583 tok/s 3,350–8,933 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	1.1 GB	Q8_0	Comfortable
4,955 tok/s 2,973–7,928 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	1.1 GB	Q8_0	Comfortable
4,955 tok/s 2,973–7,928 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	1.1 GB	Q8_0	Comfortable
4,955 tok/s 2,973–7,928 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	1.1 GB	Q8_0	Comfortable
4,701 tok/s 2,820–7,521 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	1.1 GB	Q8_0	Comfortable
4,009 tok/s 2,405–6,414 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.1 GB	Q8_0	Comfortable
4,009 tok/s 2,405–6,414 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	1.1 GB	Q8_0	Comfortable
4,009 tok/s 2,405–6,414 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	1.1 GB	Q8_0	Comfortable
4,009 tok/s 2,405–6,414 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.1 GB	Q8_0	Comfortable
4,009 tok/s 2,405–6,414 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	1.1 GB	Q8_0	Comfortable
3,052 tok/s 1,831–4,884 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	1.1 GB	Q8_0	Comfortable
3,052 tok/s 1,831–4,884 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	1.1 GB	Q8_0	Comfortable
2,544 tok/s 1,526–4,070 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	1.1 GB	Q8_0	Comfortable
2,489 tok/s 1,494–3,983 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	1.1 GB	Q8_0	Comfortable
2,434 tok/s 1,460–3,894 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	1.1 GB	Q8_0	Comfortable
2,434 tok/s 1,460–3,894 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	1.1 GB	Q8_0	Comfortable
2,434 tok/s 1,460–3,894 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	1.1 GB	Q8_0	Comfortable
2,434 tok/s 1,460–3,894 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	1.1 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,University at Buffalo
Organisation type: Academia,Academia
Country: United States of America
Published: 27 May 2022
Authors: Tri Dao, Daniel Y. Fu, Stefano Ermon, Atri Rudra, 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

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: 355M
Training data: 10,130,000,000 tokens
Batch size: 512

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: 8.9 × 10²⁰ FLOP
How it was established: Comparison with other models,Hardware

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 40 GB
Chips used: 8
Wall-clock time: 166 hours
Power draw: 6.4 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: Highly cited,Historical significance
Record confidence: Confident
Citations: 4,295

Sources

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

Reference: FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run GPT-2 Medium (FlashAttention)

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 GPT-2 Medium (FlashAttention)

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

1.1 GB

Fastest

9,544 tok/s

GPT-2 Medium (FlashAttention) is small enough at 355M 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 104 tokens per second.

At the other end, a B200 generates roughly 9,544 tokens per second on it, on the strength of 8,000 GB/s of memory bandwidth.

What this model is

GPT-2 Medium (FlashAttention) was published by Stanford University,University at Buffalo, in United States of America, in May 2022. The organisation is categorised as academia,Academia.

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.

What decides the speed

Half the cards that hold it manage more than 268.0 tokens per second, and 817 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.

Memory here is estimated from size rather than computed from the architecture, which is not recorded for this model — the numbers are indicative rather than exact.

Training and provenance

Producing it required around 8.9 × 10²⁰ FLOP of arithmetic, on NVIDIA A100 SXM4 40 GB, which is a statement about the training budget rather than about inference.

The training set ran to roughly 10,130,000,000 tokens.

It is tracked in the underlying dataset for one reason in particular: highly cited,Historical significance.

Step by step

How to choose a GPU for GPT-2 Medium (FlashAttention)

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

Look at what GPT-2 Medium (FlashAttention) actually needs — around 1.1 GB at Q8_0. No amount of processing power compensates for a card that cannot hold it.
02

Set the context length you will work at

Set the context to what you will actually use. The cache grows with the conversation, and it is the usual reason GPT-2 Medium (FlashAttention) stops fitting a card that seemed fine.
03

Set a quality floor

Compression is what makes GPT-2 Medium (FlashAttention) fit smaller cards, at some cost in accuracy — Q8_0 on the smallest card that fits. A minimum quality removes the ones that go too far.
04

Rank by throughput rather than spec sheet

Ranking by tokens per second for GPT-2 Medium (FlashAttention) follows memory bandwidth, not core counts, which is why the B200 tops it at 9,544 tok/s.
05

Look at the headroom, not just the fit

The fit column separates cards that just manage GPT-2 Medium (FlashAttention) from those with room to spare. Buy for the second if the context might grow.
06

See what else that card runs

Every card name links to its own page, which runs the same calculation across the whole model catalogue. Worth a look before buying for GPT-2 Medium (FlashAttention) alone — a card is usually bought for more than one model.

Answers

GPT-2 Medium (FlashAttention) — common questions

Would two GPUs run GPT-2 Medium (FlashAttention) faster?

Capacity adds across cards; throughput does not. Since 818 of the cards we track already hold GPT-2 Medium (FlashAttention) on their own, a second card is rarely the answer here.

Why does the quantisation differ between cards for GPT-2 Medium (FlashAttention)?

Because capacity varies, so does how hard GPT-2 Medium (FlashAttention) has to be squeezed — 1 distinct levels appear in the table above. Set a minimum quality to compare at one.

How accurate are these GPT-2 Medium (FlashAttention) speed estimates?

They are calculated from specifications rather than measured, and each carries a range — 5,727–15,271 tok/s on the B200, for instance. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

What GPU do I need to run GPT-2 Medium (FlashAttention)?

The smallest card in our catalogue that holds GPT-2 Medium (FlashAttention) is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 1.1 GB, and produces roughly 104 tokens per second. 818 cards in total can run it.

How fast is GPT-2 Medium (FlashAttention) on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 9,544 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 817 of the cards that can run GPT-2 Medium (FlashAttention) clear that.

How much VRAM does GPT-2 Medium (FlashAttention) need?

About 1.1 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.

Can I run GPT-2 Medium (FlashAttention) on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 1.1 GB and generating roughly 1,778 tokens per second — a comfortable fit.

Can I run GPT-2 Medium (FlashAttention) on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 1.1 GB and generating roughly 1,089 tokens per second — a comfortable fit.

Can I run GPT-2 Medium (FlashAttention) on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 1.1 GB and generating roughly 1,348 tokens per second — a comfortable fit.

Can I run GPT-2 Medium (FlashAttention) on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 1.1 GB and generating roughly 1,599 tokens per second — a comfortable fit.

Is GPT-2 Medium (FlashAttention) open source?

Its weights are published, so GPT-2 Medium (FlashAttention) 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 GPT-2 Medium (FlashAttention) have?

GPT-2 Medium (FlashAttention) has 355M parameters. GPT-2 Medium from here, aka GPT-2 355M: https://huggingface.co/openai-community/gpt2-medium. 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 GPT-2 Medium (FlashAttention)?

GPT-2 Medium (FlashAttention) was published by Stanford University,University at Buffalo, based in United States of America, categorised as academia,Academia.

When was GPT-2 Medium (FlashAttention) released?

GPT-2 Medium (FlashAttention) was published in May 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 GPT-2 Medium (FlashAttention) used for?

GPT-2 Medium (FlashAttention) 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 GPT-2 Medium (FlashAttention)?

The weights for GPT-2 Medium (FlashAttention) 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 GPT-2 Medium (FlashAttention)?

Around 8.9 × 10²⁰ FLOP, on NVIDIA A100 SXM4 40 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.

Can I run GPT-2 Medium (FlashAttention) if it does not fit in my GPU?

Only by offloading, which is usually a false economy: the part in system memory drags the whole thing down. Our figures for GPT-2 Medium (FlashAttention) assume it is fully resident.

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