base LM+GNN+kNN TPS calculator

Open weights Shannon.AI,Nanjing University,Nanyang Technological University,Zhejiang University (ZJU) 274M 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 · 135 tok/s

Fastest card

B200

12,366 tok/s · 180 GB

Which GPUs can run base LM+GNN+kNN?

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
12,366 tok/s 7,419–19,785 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	1.0 GB	Q8_0	Comfortable
12,366 tok/s 7,419–19,785 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	1.0 GB	Q8_0	Comfortable
9,874 tok/s 5,925–15,799 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	1.0 GB	Q8_0	Comfortable
9,874 tok/s 5,925–15,799 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	1.0 GB	Q8_0	Comfortable
7,897 tok/s 4,738–12,635 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	1.0 GB	Q8_0	Comfortable
7,559 tok/s 4,535–12,094 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.0 GB	Q8_0	Comfortable
7,559 tok/s 4,535–12,094 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.0 GB	Q8_0	Comfortable
7,234 tok/s 4,340–11,574 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	1.0 GB	Q8_0	Comfortable
6,420 tok/s 3,852–10,272 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	1.0 GB	Q8_0	Comfortable
6,420 tok/s 3,852–10,272 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	1.0 GB	Q8_0	Comfortable
6,420 tok/s 3,852–10,272 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	1.0 GB	Q8_0	Comfortable
6,090 tok/s 3,654–9,744 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
5,194 tok/s 3,116–8,310 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
5,194 tok/s 3,116–8,310 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	1.0 GB	Q8_0	Comfortable
5,194 tok/s 3,116–8,310 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
5,194 tok/s 3,116–8,310 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
5,194 tok/s 3,116–8,310 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	1.0 GB	Q8_0	Comfortable
3,955 tok/s 2,373–6,327 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	1.0 GB	Q8_0	Comfortable
3,955 tok/s 2,373–6,327 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	1.0 GB	Q8_0	Comfortable
3,295 tok/s 1,977–5,273 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	1.0 GB	Q8_0	Comfortable
3,225 tok/s 1,935–5,160 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	1.0 GB	Q8_0	Comfortable
3,153 tok/s 1,892–5,045 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	1.0 GB	Q8_0	Comfortable
3,153 tok/s 1,892–5,045 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	1.0 GB	Q8_0	Comfortable
3,153 tok/s 1,892–5,045 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	1.0 GB	Q8_0	Comfortable
3,153 tok/s 1,892–5,045 · 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: Shannon.AI,Nanjing University,Nanyang Technological University,Zhejiang University (ZJU)
Organisation type: Industry,Academia,Academia,Academia
Country: China, Singapore
Published: 17 October 2021
Authors: Yuxian Meng, Shi Zong, Xiaoya Li, Xiaofei Sun, Tianwei Zhang, Fei Wu, Jiwei Li

What it does

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

Domain: Language
Task: Language modeling
Base model: Transformer (Adaptive Input Embeddings) WT103
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: 274M
Training data: 103,000,000 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: 5.3 × 10¹⁹ FLOP
How it was established: Operation counting

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: Confident
Citations: 46
Benchmark data: base LM+GNN+kNN

Sources

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

Reference: GNN-LM: Language Modeling based on Global Contexts via GNN
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run base LM+GNN+kNN

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 base LM+GNN+kNN

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

The hardware side

Minimum card

Tesla C1080

Memory needed

1.0 GB

Fastest

12,366 tok/s

base LM+GNN+kNN is small enough at 274M parameters that hardware is rarely the obstacle — 818 of the cards we track can run it, including cards several years old.

At the low end, a Tesla C1080 handles it — 4 GB, at Q8_0, for about 135 tokens per second.

Top of the range is the B200, at roughly 12,366 tokens per second thanks to 8,000 GB/s of bandwidth.

Where it came from

base LM+GNN+kNN was published by Shannon.AI,Nanjing University,Nanyang Technological University,Zhejiang University (ZJU), in China, in October 2021. The organisation is categorised as industry,Academia,Academia,Academia.

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

It builds on Transformer (Adaptive Input Embeddings) WT103, which is why it shares that model's general shape and size.

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

Understanding the speeds

The median result is around 347.2 tokens per second; 818 cards produce text faster than most people read it.

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

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

The training run consumed about 5.3 × 10¹⁹ FLOP. 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 base LM+GNN+kNN

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

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

Set the context length you will work at

Longer conversations cost memory on top of what the weights need. Move the slider to your real working length before trusting any row for base LM+GNN+kNN.
03

Choose how far you will compress it

Compression is what makes base LM+GNN+kNN 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

Ranking by tokens per second for base LM+GNN+kNN follows memory bandwidth, not core counts, which is why the B200 tops it at 12,366 tok/s.
05

Check the fit verdict before buying

The fit column separates cards that just manage base LM+GNN+kNN from those with room to spare. Buy for the second if the context might grow.
06

Open the card you have settled on

Every card name links to its own page, which runs the same calculation across the whole model catalogue. Worth a look before buying for base LM+GNN+kNN alone — a card is usually bought for more than one model.

Answers

base LM+GNN+kNN — common questions

Where can I download base LM+GNN+kNN?

The weights for base LM+GNN+kNN 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 base LM+GNN+kNN?

Around 5.3 × 10¹⁹ 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.

Can I run base LM+GNN+kNN 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 base LM+GNN+kNN is rarely worth using. Every figure here assumes the whole model is on the card.

Would two GPUs run base LM+GNN+kNN faster?

Capacity adds across cards; throughput does not. Since 818 of the cards we track already hold base LM+GNN+kNN on their own, a second card is rarely the answer here.

Why does the quantisation differ between cards for base LM+GNN+kNN?

Because capacity varies, so does how hard base LM+GNN+kNN has to be squeezed — 1 distinct levels appear in the table above. Set a minimum quality to compare at one.

How accurate are these base LM+GNN+kNN 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 7,419–19,785 tok/s on the B200 rather than a single number.

What GPU do I need to run base LM+GNN+kNN?

The smallest card in our catalogue that holds base LM+GNN+kNN is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 1.0 GB, and produces roughly 135 tokens per second. 818 cards in total can run it.

How fast is base LM+GNN+kNN on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 12,366 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 base LM+GNN+kNN clear that.

How much VRAM does base LM+GNN+kNN 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.

Can I run base LM+GNN+kNN 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 2,303 tokens per second — a comfortable fit.

Can I run base LM+GNN+kNN 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 1,410 tokens per second — a comfortable fit.

Can I run base LM+GNN+kNN 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 1,747 tokens per second — a comfortable fit.

Can I run base LM+GNN+kNN 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 2,071 tokens per second — a comfortable fit.

Is base LM+GNN+kNN open source?

Its weights are published, so base LM+GNN+kNN 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 base LM+GNN+kNN have?

base LM+GNN+kNN has 274M parameters. 274M (table 1) "We use the base version of deep Transformer language model with adaptive embeddings (Baevski & Auli, 2018) as our base LM. This model has 16 decoder layers. The dimensionality of word representations is 1,024, the number of multi-attention heads is 16, and the inner dimensionality of feedforward layers is 4,096.". 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 base LM+GNN+kNN?

base LM+GNN+kNN was published by Shannon.AI,Nanjing University,Nanyang Technological University,Zhejiang University (ZJU), based in China, categorised as industry,Academia,Academia,Academia.

When was base LM+GNN+kNN released?

base LM+GNN+kNN 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 base LM+GNN+kNN used for?

base LM+GNN+kNN 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.

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