Qwen2.5-72B TPS calculator

Open weights Alibaba 72.7B parameters September 2024

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

61 cards that can run it

818 cards we hold specifications for

Smallest card that fits

A100 PCIe 40 GB

40 GB · Q3_K_M · 24.5 tok/s

Fastest card

B200

46.6 tok/s · 180 GB

Which GPUs can run Qwen2.5-72B?

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

61 cards match

Calculating

					Needs	Quantisation	Fit
46.6 tok/s 40–56	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	76.5 GB	Q8_0	Comfortable
46.6 tok/s 40–56	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	76.5 GB	Q8_0	Comfortable
37.2 tok/s 22–60 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	76.5 GB	Q8_0	Comfortable
37.2 tok/s 22–60 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	76.5 GB	Q8_0	Comfortable
29.8 tok/s 18–48 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	76.5 GB	Q8_0	Comfortable
28.5 tok/s 24–34	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	76.5 GB	Q8_0	Comfortable
28.5 tok/s 24–34	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	76.5 GB	Q8_0	Comfortable
28.4 tok/s 24–34	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	59.6 GB	Q6_K	Comfortable
28.4 tok/s 24–34	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	59.6 GB	Q6_K	Comfortable
27.3 tok/s 16–44 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	76.5 GB	Q8_0	Comfortable
25.2 tok/s 21–30	GRID A100B NVIDIA	48 GB	1,870 GB/s	May 2020	42.6 GB	Q4_K_M	Tight
24.5 tok/s 21–29	A100 PCIe 40 GB NVIDIA	40 GB	1,560 GB/s	Jun 2020	34.2 GB	Q3_K_M	Tight
24.5 tok/s 21–29	A100 SXM4 40 GB NVIDIA	40 GB	1,560 GB/s	May 2020	34.2 GB	Q3_K_M	Tight
24.5 tok/s 21–29	A800 PCIe 40 GB NVIDIA	40 GB	1,560 GB/s	Nov 2022	34.2 GB	Q3_K_M	Tight
24.2 tok/s 15–39 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	76.5 GB	Q8_0	Comfortable
24.2 tok/s 15–39 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	76.5 GB	Q8_0	Comfortable
24.2 tok/s 15–39 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	76.5 GB	Q8_0	Comfortable
23.0 tok/s 20–28	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	76.5 GB	Q8_0	Tight
21.0 tok/s 18–25	H100 SXM5 64 GB NVIDIA	64 GB	2,020 GB/s	Mar 2023	51.1 GB	Q5_K_M	Tight
19.6 tok/s 17–23	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	76.5 GB	Q8_0	Tight
19.6 tok/s 17–23	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	76.5 GB	Q8_0	Tight
19.6 tok/s 17–23	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	76.5 GB	Q8_0	Tight
18.0 tok/s 15–22	RTX PRO 5000 Blackwell NVIDIA	48 GB	1,340 GB/s	Mar 2025	42.6 GB	Q4_K_M	Tight
17.3 tok/s 15–21	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	59.6 GB	Q6_K	Comfortable
17.3 tok/s 15–21	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	59.6 GB	Q6_K	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: Alibaba
Organisation type: Industry
Country: China
Published: 19 September 2024
Authors: Qwen Team

What it does

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

Domain: Language
Task: Language modeling/generation, Question answering, Quantitative reasoning
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: 72.7B
Training data: 18,000,000,000,000 tokens
Epochs: 1

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 × 10²⁴ FLOP
How it was established: Operation counting

The training run

What it physically took to train: which chips, how many, for how long, and what that drew from the wall.

Data centre: There is no paper to reference, no information about hardware used for training found in media.

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
Hugging Face: Qwen

How it is classified

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

Likely above 10²³ FLOP: Yes
Why it is tracked: Training cost
Record confidence: Confident

Sources

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

Reference: Qwen2.5: A Party of Foundation Models!
Last updated: 18 December 2025

The extremes

The ten fastest GPUs that run Qwen2.5-72B

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 Qwen2.5-72B

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

Hardware requirements in practice

Minimum card

A100 PCIe 40 GB

Memory needed

34.2 GB

Fastest

46.6 tok/s

Qwen2.5-72B reaches a parameter count of 72.7B. That puts it above consumer hardware, into the range where a card is bought for this purpose rather than repurposed for it. The number of cards we track that can hold it: 61.

The entry point is A100 PCIe 40 GB, with a memory capacity of 40 GB, running it at a compression of Q3_K_M and producing around 24.5 tokens per second.

Top of the range is B200, generating roughly 46.6 tokens per second on the strength of a memory bandwidth of 8,000 GB/s.

About this model

Qwen2.5-72B was published by Alibaba, in the country recorded as China, during September 2024. It comes out of an organisation categorised as industry.

It works in the domain of Language, and is recorded as performing the task of language modeling/generation, Question answering, Quantitative reasoning.

The weights are published, so it can be downloaded and run on your own hardware indefinitely, offline, with no account attached. On Hugging Face it is published under the organisation Qwen.

How fast it runs, and why

Across every card that can run it, the middle of the range sits at 16.4 tokens per second. Producing text faster than most people read it: 50 of them.

Every weight participates in every token here, so bandwidth is the whole story: the ranking below is effectively a ranking of memory throughput.

Its attention layout is on file, so the memory figures are computed exactly rather than approximated.

How it was trained

The training run consumed about 7.8 × 10²⁴ FLOP. 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 18,000,000,000,000 tokens of text.

It is tracked in the underlying dataset for one reason in particular: training cost.

Step by step

How to choose a GPU for Qwen2.5-72B

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

01

Start from the memory column

Start from what it actually needs, which is the requirement of Qwen2.5-72B, needing around 34.2 GB at a compression of Q3_K_M. Capacity is the gate — a card either holds it or it does not.
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 a card that seemed fine stops fitting Qwen2.5-72B.
03

Choose how far you will compress it

Each card runs the least-compressed copy it can hold, reaching a compression of Q3_K_M on the smallest card that fits. 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

The speed ordering is effectively an ordering by memory bandwidth, for Qwen2.5-72B. It will not match a gaming ordering, because generation is bound by memory bandwidth. The card topping the list is B200, at 46.6 tok/s.
05

Check the fit verdict before buying

The fit column separates cards that just manage it from those with room to spare, in the case of Qwen2.5-72B. 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

Open the card you have settled on

Each card page repeats this sweep for every model we hold, answering what else the hardware is good for beyond Qwen2.5-72B.

Answers

Qwen2.5-72B — common questions

Qwen2.5-72B— who created it?

It was published by Alibaba, based in China, an organisation categorised as industry.

Qwen2.5-72B— when was it released?

It was published in September 2024. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

Qwen2.5-72B— what is it used for?

It works in the domain of Language, and is recorded as handling the task of language modeling/generation, Question answering, Quantitative reasoning. 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.

Qwen2.5-72B— where can I download it?

Its weights are published on Hugging Face, under the organisation Qwen. We do not host model files — this site calculates what hardware is needed to run them.

Qwen2.5-72B— how much compute was used to train it?

Training consumed around 7.8 × 10²⁴ FLOP. 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.

Qwen2.5-72B— can I run it if it does not fit in my GPU?

Only by offloading, which is usually a false economy: the part held in system memory drags the whole thing down. The nearest miss we calculate falls short by 13.8 GB. Every figure here assumes the whole model is resident on the card.

Qwen2.5-72B— would two GPUs run it faster?

Capacity adds across cards; throughput does not. The number of cards already holding it on their own: 61. So a second card is rarely the answer here.

Qwen2.5-72B— 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: 5. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.

Qwen2.5-72B— 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: 40–56 tok/s on B200. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

Qwen2.5-72B— what GPU do I need to run it?

The smallest card in our catalogue that holds it is A100 PCIe 40 GB, with a memory capacity of 40 GB. It runs the model at a compression of Q3_K_M using about 34.2 GB, and produces roughly 24.5 tokens per second. The number of cards able to run it in total: 61.

Qwen2.5-72B— how fast is it on a GPU?

It depends on the card. The quickest we calculate is B200, at about 46.6 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and the number of cards clearing that: 50.

Qwen2.5-72B— how much VRAM does it need?

It needs about 34.2 GB at a compression of Q3_K_M, 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.

Qwen2.5-72B— 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.

Qwen2.5-72B— how many parameters does it have?

It has a parameter count of 72.7B. 72.7B. 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.

Source

Original publication

Record last updated 18 December 2025

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