ERNIE 3.0 TPS calculator

Open weights Baidu 10B parameters July 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

509 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Xeon Phi 5110P

8 GB · Q4_K_M · 20.3 tok/s

Fastest card

B200

339 tok/s · 180 GB

Which GPUs can run ERNIE 3.0?

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

509 cards match

Calculating

					Needs	Quantisation	Fit
339 tok/s 203–542 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	11.4 GB	Q8_0	Comfortable
339 tok/s 203–542 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	11.4 GB	Q8_0	Comfortable
271 tok/s 162–433 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	11.4 GB	Q8_0	Comfortable
271 tok/s 162–433 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	11.4 GB	Q8_0	Comfortable
216 tok/s 130–346 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	11.4 GB	Q8_0	Comfortable
207 tok/s 124–331 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	11.4 GB	Q8_0	Comfortable
207 tok/s 124–331 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	11.4 GB	Q8_0	Comfortable
198 tok/s 119–317 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	11.4 GB	Q8_0	Comfortable
176 tok/s 106–281 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	11.4 GB	Q8_0	Comfortable
176 tok/s 106–281 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	11.4 GB	Q8_0	Comfortable
176 tok/s 106–281 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	11.4 GB	Q8_0	Comfortable
167 tok/s 100–267 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	11.4 GB	Q8_0	Comfortable
146 tok/s 87–233 · low confidence	CMP 170HX 8 GB NVIDIA	8 GB	1,490 GB/s	Sep 2021	6.7 GB	Q4_K_M	Tight
142 tok/s 85–228 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	11.4 GB	Q8_0	Comfortable
142 tok/s 85–228 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	11.4 GB	Q8_0	Comfortable
142 tok/s 85–228 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	11.4 GB	Q8_0	Comfortable
142 tok/s 85–228 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	11.4 GB	Q8_0	Comfortable
142 tok/s 85–228 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	11.4 GB	Q8_0	Comfortable
118 tok/s 71–189 · low confidence	CMP 170HX 10 GB NVIDIA	10 GB	1,560 GB/s	Sep 2021	7.9 GB	Q5_K_M	Tight
108 tok/s 65–173 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	11.4 GB	Q8_0	Comfortable
108 tok/s 65–173 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	11.4 GB	Q8_0	Comfortable
90.3 tok/s 54–144 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	11.4 GB	Q8_0	Comfortable
88.4 tok/s 53–141 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	11.4 GB	Q8_0	Comfortable
86.4 tok/s 52–138 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	11.4 GB	Q8_0	Comfortable
86.4 tok/s 52–138 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	11.4 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: Baidu
Organisation type: Industry
Country: China
Published: 5 July 2021
Authors: Y Sun, S Wang, S Feng, S Ding, C Pang

What it does

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

Domain: Language
Task: Language modeling, Language modeling/generation, Text classification, Question answering
Approach: Self-supervised learning
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: 10B
Training data: 375,000,000,000 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: 2.3 × 10²² 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.

Training hardware: NVIDIA V100
Chips used: 384
Power draw: 232.9 kW
Compute cost: $39,104

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: 590

Sources

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

Reference: ERNIE 3.0: Large-scale Knowledge Enhanced Pre-training for Language Understanding and Generation
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run ERNIE 3.0

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 ERNIE 3.0

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

Xeon Phi 5110P

Memory needed

6.7 GB

Fastest

339 tok/s

ERNIE 3.0 is small enough at 10B parameters that hardware is rarely the obstacle — 509 of the cards we track can run it, including cards several years old.

The entry point is the Xeon Phi 5110P: 8 GB of memory, Q4_K_M compression, roughly 20.3 tokens per second.

At the other end, a B200 generates roughly 339 tokens per second on it, on the strength of 8,000 GB/s of memory bandwidth.

Background

ERNIE 3.0 was published by Baidu, in China, in July 2021. The organisation is categorised as industry.

It works in Language, and is recorded as doing language modeling, Language modeling/generation, Text classification, Question answering.

The weights being open is what puts this page in the calculator rather than only in the catalogue: it is a model you can actually hold.

Reading the throughput figures

Across every card that can run it, the middle of the range is about 22.0 tokens per second, and 469 of them clear the ten tokens per second that roughly matches reading speed.

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.

What went into building it

Producing it required around 2.3 × 10²² FLOP of arithmetic, on NVIDIA V100, which is a statement about the training budget rather than about inference.

It was trained on about 375,000,000,000 tokens of text.

Its inclusion criterion is sOTA improvement.

Step by step

How to choose a GPU for ERNIE 3.0

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

Every card here has been checked against ERNIE 3.0 — around 6.7 GB at Q4_K_M. Capacity is the gate — a card either holds it or it does not.
02

Decide how long your conversations run

Longer conversations cost memory on top of what the weights need. Move the slider to your real working length before trusting any row for ERNIE 3.0.
03

Decide how much compression you will accept

The quantisation column varies by card, because a bigger card holds a more accurate copy of ERNIE 3.0 — Q4_K_M on the smallest card that fits. Set a floor to hold the comparison at one level.
04

Rank by throughput rather than spec sheet

Sort by speed to see how cards rank for ERNIE 3.0. It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 339 tok/s.
05

Look at the headroom, not just the fit

The fit column separates cards that just manage ERNIE 3.0 from those with room to spare. Buy for the second if the context might grow.
06

Check the card from the other side

Every card name links to its own page, which runs the same calculation across the whole model catalogue. Worth a look before buying for ERNIE 3.0 alone — a card is usually bought for more than one model.

Answers

ERNIE 3.0 — common questions

Can I run ERNIE 3.0 on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 11.4 GB and generating roughly 56.8 tokens per second — a comfortable fit.

Is ERNIE 3.0 open source?

Its weights are published, so ERNIE 3.0 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 ERNIE 3.0 have?

ERNIE 3.0 has 10B parameters. "We trained the model with 10 billion parameters on a 4TB corpus consisting of plain texts and a large-scale knowledge graph.". 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 ERNIE 3.0?

ERNIE 3.0 was published by Baidu, based in China, categorised as industry.

When was ERNIE 3.0 released?

ERNIE 3.0 was published in July 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 ERNIE 3.0 used for?

ERNIE 3.0 works in Language, and is recorded as handling language modeling, Language modeling/generation, Text classification, Question answering. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Where can I download ERNIE 3.0?

The weights for ERNIE 3.0 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 ERNIE 3.0?

Around 2.3 × 10²² 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 ERNIE 3.0 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 ERNIE 3.0 is rarely worth using — the nearest miss we calculate is short by 1.3 GB. Every figure here assumes the whole model is on the card.

Would two GPUs run ERNIE 3.0 faster?

Two cards buy memory rather than speed. That matters for ERNIE 3.0 only if one card cannot hold it — 509 can, so a second adds little.

Why does the quantisation differ between cards for ERNIE 3.0?

A larger card holds a more accurate copy. Across the cards that run ERNIE 3.0, 4 compression levels are used; the floor control above pins it to one.

How accurate are these ERNIE 3.0 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 203–542 tok/s on the B200 rather than a single number.

What GPU do I need to run ERNIE 3.0?

The smallest card in our catalogue that holds ERNIE 3.0 is the Xeon Phi 5110P, with 8 GB of memory. It runs the model at Q4_K_M using about 6.7 GB, and produces roughly 20.3 tokens per second. 509 cards in total can run it.

How fast is ERNIE 3.0 on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 339 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 469 of the cards that can run ERNIE 3.0 clear that.

How much VRAM does ERNIE 3.0 need?

About 6.7 GB at Q4_K_M 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 ERNIE 3.0 on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q4_K_M, using about 6.7 GB and generating roughly 146 tokens per second — a tight fit.

Can I run ERNIE 3.0 on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q6_K, using about 9.1 GB and generating roughly 56.2 tokens per second — a tight fit.

Can I run ERNIE 3.0 on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 11.4 GB and generating roughly 47.9 tokens per second — a comfortable fit.

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