CPM-Large TPS calculator

Open weights Tsinghua University,Beijing Academy of Artificial Intelligence / BAAI 2.6B parameters December 2020

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 · 14.2 tok/s

Fastest card

B200

1,303 tok/s · 180 GB

Which GPUs can run CPM-Large?

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
1,303 tok/s 782–2,085 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	3.5 GB	Q8_0	Comfortable
1,303 tok/s 782–2,085 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	3.5 GB	Q8_0	Comfortable
1,041 tok/s 624–1,665 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	3.5 GB	Q8_0	Comfortable
1,041 tok/s 624–1,665 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	3.5 GB	Q8_0	Comfortable
832 tok/s 499–1,332 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	3.5 GB	Q8_0	Comfortable
797 tok/s 478–1,275 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	3.5 GB	Q8_0	Comfortable
797 tok/s 478–1,275 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	3.5 GB	Q8_0	Comfortable
762 tok/s 457–1,220 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	3.5 GB	Q8_0	Comfortable
677 tok/s 406–1,083 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	3.5 GB	Q8_0	Comfortable
677 tok/s 406–1,083 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	3.5 GB	Q8_0	Comfortable
677 tok/s 406–1,083 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	3.5 GB	Q8_0	Comfortable
642 tok/s 385–1,027 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	3.5 GB	Q8_0	Comfortable
547 tok/s 328–876 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	3.5 GB	Q8_0	Comfortable
547 tok/s 328–876 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	3.5 GB	Q8_0	Comfortable
547 tok/s 328–876 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	3.5 GB	Q8_0	Comfortable
547 tok/s 328–876 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	3.5 GB	Q8_0	Comfortable
547 tok/s 328–876 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	3.5 GB	Q8_0	Comfortable
417 tok/s 250–667 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	3.5 GB	Q8_0	Comfortable
417 tok/s 250–667 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	3.5 GB	Q8_0	Comfortable
347 tok/s 208–556 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	3.5 GB	Q8_0	Comfortable
340 tok/s 204–544 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	3.5 GB	Q8_0	Comfortable
332 tok/s 199–532 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	3.5 GB	Q8_0	Comfortable
332 tok/s 199–532 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	3.5 GB	Q8_0	Comfortable
332 tok/s 199–532 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	3.5 GB	Q8_0	Comfortable
332 tok/s 199–532 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	3.5 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: Tsinghua University,Beijing Academy of Artificial Intelligence / BAAI
Organisation type: Academia,Academia
Country: China
Published: 1 December 2020
Authors: Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun

What it does

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

Domain: Language
Task: Language modeling
Approach: Self-supervised learning

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: 2.6B
Training data: 16,700,000,000 tokens

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: 2.6 × 10²⁰ FLOP
How it was established: Third-party estimation

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 V100
Chips used: 64
Wall-clock time: 336 hours (14 days)
Power draw: 39.0 kW
Compute cost: $7,340

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

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: Confident
Citations: 129

Sources

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

Reference: CPM: A Large-scale Generative Chinese Pre-trained Language Model
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run CPM-Large

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 CPM-Large

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 it takes to run this model

Minimum card

Tesla C1080

Memory needed

3.5 GB

Fastest

1,303 tok/s

CPM-Large is small enough at 2.6B parameters that hardware is rarely the obstacle — 818 of the cards we track can run it, including cards several years old.

The smallest card that holds it is the Tesla C1080 with 4 GB, running it at Q8_0 and producing around 14.2 tokens per second.

The quickest result comes from a B200 at around 1,303 tokens per second — its 8,000 GB/s of bandwidth is what buys that.

Background

CPM-Large was published by Tsinghua University,Beijing Academy of Artificial Intelligence / BAAI, in China, in December 2020. academia,Academia is the category the publisher falls under.

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

Because its weights were released, nothing about running it depends on a provider staying available — it is yours once downloaded.

Reading the throughput figures

Half the cards that hold it manage more than 36.6 tokens per second, and 778 exceed reading speed outright.

It is a dense model, so every parameter is read for every token produced. That makes speed track memory bandwidth almost exactly — a card with twice the bandwidth generates roughly twice as fast.

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.

What went into building it

Training it took roughly 2.6 × 10²⁰ FLOP of computation, on NVIDIA V100 — a measure of what producing the model cost, not of how fast it answers.

The training set ran to roughly 16,700,000,000 tokens.

Its inclusion criterion is sOTA improvement.

Step by step

How to choose a GPU for CPM-Large

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

Every card here has been checked against CPM-Large — around 3.5 GB at Q8_0. Capacity is the gate — a card either holds it or it does not.
02

Match the context to your actual use

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

Set a quality floor

Compression is what makes CPM-Large 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

Compare tokens per second, not specifications

The speed ordering for CPM-Large is effectively an ordering by memory bandwidth, which is why the B200 tops it at 1,303 tok/s.
05

Look at the headroom, not just the fit

A tight fit runs CPM-Large but leaves nothing spare for a longer conversation; comfortable has headroom. If you expect to grow the context, buy for comfortable.
06

Check the card from the other side

Following a card through to its own page shows every other model it can hold, which is the question that follows once CPM-Large is settled.

Answers

CPM-Large — common questions

Who created CPM-Large?

CPM-Large was published by Tsinghua University,Beijing Academy of Artificial Intelligence / BAAI, based in China, categorised as academia,Academia.

When was CPM-Large released?

CPM-Large was published in December 2020. 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 CPM-Large used for?

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

Where can I download CPM-Large?

The weights for CPM-Large 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 CPM-Large?

Around 2.6 × 10²⁰ FLOP, on NVIDIA V100. 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 CPM-Large if it does not fit in my GPU?

It can be split between the card and system memory, but CPM-Large generates painfully slowly that way. Nothing on this page assumes offloading.

Would two GPUs run CPM-Large faster?

Capacity adds across cards; throughput does not. Since 818 of the cards we track already hold CPM-Large on their own, a second card is rarely the answer here.

Why does the quantisation differ between cards for CPM-Large?

Because capacity varies, so does how hard CPM-Large has to be squeezed — 1 distinct levels appear in the table above. Set a minimum quality to compare at one.

How accurate are these CPM-Large speed estimates?

Every figure is derived from memory bandwidth and model size, not benchmarked. That is why each is published as a range such as 782–2,085 tok/s on the B200 rather than a single number.

What GPU do I need to run CPM-Large?

The smallest card in our catalogue that holds CPM-Large is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 3.5 GB, and produces roughly 14.2 tokens per second. 818 cards in total can run it.

How fast is CPM-Large on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 1,303 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 778 of the cards that can run CPM-Large clear that.

How much VRAM does CPM-Large need?

About 3.5 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 CPM-Large on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 3.5 GB and generating roughly 243 tokens per second — a comfortable fit.

Can I run CPM-Large on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 3.5 GB and generating roughly 149 tokens per second — a comfortable fit.

Can I run CPM-Large on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 3.5 GB and generating roughly 184 tokens per second — a comfortable fit.

Can I run CPM-Large on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 3.5 GB and generating roughly 218 tokens per second — a comfortable fit.

Is CPM-Large open source?

Its weights are published, so CPM-Large 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 CPM-Large have?

CPM-Large has 2.6B parameters. "To the best of our knowledge, CPM, with 2.6 billion parameters and 100GB Chinese training data, is the largest Chinese pre-trained language mode". 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 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