EVA-CLIP (EVA-02-CLIP-E/14+) TPS calculator

Open weights Beijing Academy of Artificial Intelligence / BAAI,Huazhong University of Science and Technology 5B parameters March 2023

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 · IQ4_XS · 18.1 tok/s

Fastest card

B200

678 tok/s · 180 GB

Which GPUs can run EVA-CLIP (EVA-02-CLIP-E/14+)?

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
678 tok/s 407–1,084 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	6.1 GB	Q8_0	Comfortable
678 tok/s 407–1,084 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	6.1 GB	Q8_0	Comfortable
541 tok/s 325–866 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	6.1 GB	Q8_0	Comfortable
541 tok/s 325–866 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	6.1 GB	Q8_0	Comfortable
433 tok/s 260–692 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	6.1 GB	Q8_0	Comfortable
414 tok/s 249–663 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	6.1 GB	Q8_0	Comfortable
414 tok/s 249–663 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	6.1 GB	Q8_0	Comfortable
396 tok/s 238–634 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	6.1 GB	Q8_0	Comfortable
352 tok/s 211–563 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	6.1 GB	Q8_0	Comfortable
352 tok/s 211–563 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	6.1 GB	Q8_0	Comfortable
352 tok/s 211–563 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	6.1 GB	Q8_0	Comfortable
334 tok/s 200–534 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	6.1 GB	Q8_0	Comfortable
285 tok/s 171–455 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	6.1 GB	Q8_0	Comfortable
285 tok/s 171–455 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	6.1 GB	Q8_0	Comfortable
285 tok/s 171–455 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	6.1 GB	Q8_0	Comfortable
285 tok/s 171–455 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	6.1 GB	Q8_0	Comfortable
285 tok/s 171–455 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	6.1 GB	Q8_0	Comfortable
217 tok/s 130–347 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	6.1 GB	Q8_0	Comfortable
217 tok/s 130–347 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	6.1 GB	Q8_0	Comfortable
181 tok/s 108–289 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	6.1 GB	Q8_0	Comfortable
177 tok/s 106–283 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	6.1 GB	Q8_0	Comfortable
173 tok/s 104–276 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	6.1 GB	Q8_0	Comfortable
173 tok/s 104–276 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	6.1 GB	Q8_0	Comfortable
173 tok/s 104–276 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	6.1 GB	Q8_0	Comfortable
173 tok/s 104–276 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	6.1 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: Beijing Academy of Artificial Intelligence / BAAI,Huazhong University of Science and Technology
Organisation type: Academia,Academia
Country: China
Published: 27 March 2023
Authors: Quan Sun, Yuxin Fang, Ledell Wu, Xinlong Wang, Yue Cao

What it does

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

Domain: Vision
Task: Image classification
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: 5B
Training data: tokens
Batch size: 144,000

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: 3.5 × 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 RTX A1000
Chips used: 144
Power draw: 17.2 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: Unreleased
Hugging Face: QuanSun

How it is classified

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

Record confidence: Confident

Sources

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

Reference: EVA-CLIP: Improved Training Techniques for CLIP at Scale
Last updated: 11 February 2026

The extremes

The ten fastest GPUs that run EVA-CLIP (EVA-02-CLIP-E/14+)

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 EVA-CLIP (EVA-02-CLIP-E/14+)

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

3.4 GB

Fastest

678 tok/s

EVA-CLIP (EVA-02-CLIP-E/14+) is small enough at 5B 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 IQ4_XS, for about 18.1 tokens per second.

A B200 is the fastest we calculate for it: about 678 tokens per second, from 8,000 GB/s of memory bandwidth.

About this model

EVA-CLIP (EVA-02-CLIP-E/14+) was published by Beijing Academy of Artificial Intelligence / BAAI,Huazhong University of Science and Technology, in China, in March 2023. academia,Academia is the category the publisher falls under.

It works in Vision, and is recorded as doing image classification.

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

How fast it runs, and why

The median result is around 27.1 tokens per second; 771 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.

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.

Training and provenance

The training run consumed about 3.5 × 10²² FLOP, on NVIDIA RTX A1000. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

Step by step

How to choose a GPU for EVA-CLIP (EVA-02-CLIP-E/14+)

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 EVA-CLIP (EVA-02-CLIP-E/14+) actually needs — around 3.4 GB at IQ4_XS. No amount of processing power compensates for a card that cannot hold it.
02

Set the context length you will work at

The conversation occupies memory too, and grows as it goes. Set the slider to the length you expect: at long context EVA-CLIP (EVA-02-CLIP-E/14+) can slip off a card that handles short questions easily.
03

Set a quality floor

Each card runs the least-compressed copy it can hold — IQ4_XS on the smallest card that fits. Setting a floor drops the cards that only manage EVA-CLIP (EVA-02-CLIP-E/14+) by squeezing it further than you would want.
04

Compare tokens per second, not specifications

Sort by speed to see how cards rank for EVA-CLIP (EVA-02-CLIP-E/14+). It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 678 tok/s.
05

Look at the headroom, not just the fit

Tight means EVA-CLIP (EVA-02-CLIP-E/14+) loads and works, with no room to raise the context later. Comfortable means you can. The difference matters more than a few tokens per second.
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 EVA-CLIP (EVA-02-CLIP-E/14+) alone — a card is usually bought for more than one model.

Answers

EVA-CLIP (EVA-02-CLIP-E/14+) — common questions

Where can I download EVA-CLIP (EVA-02-CLIP-E/14+)?

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

How much compute was used to train EVA-CLIP (EVA-02-CLIP-E/14+)?

Around 3.5 × 10²² FLOP, on NVIDIA RTX A1000. 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 EVA-CLIP (EVA-02-CLIP-E/14+) 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 EVA-CLIP (EVA-02-CLIP-E/14+) assume it is fully resident.

Would two GPUs run EVA-CLIP (EVA-02-CLIP-E/14+) faster?

A second card roughly doubles the memory available but not the generation rate. With 818 cards already able to run EVA-CLIP (EVA-02-CLIP-E/14+) alone, the case for pairing is weak.

Why does the quantisation differ between cards for EVA-CLIP (EVA-02-CLIP-E/14+)?

A larger card holds a more accurate copy. Across the cards that run EVA-CLIP (EVA-02-CLIP-E/14+), 4 compression levels are used; the floor control above pins it to one.

How accurate are these EVA-CLIP (EVA-02-CLIP-E/14+) speed estimates?

These are estimates with real error bars. The fastest result here, 407–1,084 tok/s on the B200, could reasonably land anywhere in its published range depending on which runtime you use.

What GPU do I need to run EVA-CLIP (EVA-02-CLIP-E/14+)?

The smallest card in our catalogue that holds EVA-CLIP (EVA-02-CLIP-E/14+) is the Tesla C1080, with 4 GB of memory. It runs the model at IQ4_XS using about 3.4 GB, and produces roughly 18.1 tokens per second. 818 cards in total can run it.

How fast is EVA-CLIP (EVA-02-CLIP-E/14+) on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 678 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 771 of the cards that can run EVA-CLIP (EVA-02-CLIP-E/14+) clear that.

How much VRAM does EVA-CLIP (EVA-02-CLIP-E/14+) need?

About 3.4 GB at IQ4_XS 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 EVA-CLIP (EVA-02-CLIP-E/14+) on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 6.1 GB and generating roughly 126 tokens per second — a tight fit.

Can I run EVA-CLIP (EVA-02-CLIP-E/14+) on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 6.1 GB and generating roughly 77.3 tokens per second — a comfortable fit.

Can I run EVA-CLIP (EVA-02-CLIP-E/14+) on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 6.1 GB and generating roughly 95.7 tokens per second — a comfortable fit.

Can I run EVA-CLIP (EVA-02-CLIP-E/14+) on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 6.1 GB and generating roughly 114 tokens per second — a comfortable fit.

Is EVA-CLIP (EVA-02-CLIP-E/14+) open source?

Its weights are published, so EVA-CLIP (EVA-02-CLIP-E/14+) 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 EVA-CLIP (EVA-02-CLIP-E/14+) have?

EVA-CLIP (EVA-02-CLIP-E/14+) has 5B parameters. 5b (table 1(a)) image parameters: 4.4B text parameters: 695M. 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 EVA-CLIP (EVA-02-CLIP-E/14+)?

EVA-CLIP (EVA-02-CLIP-E/14+) was published by Beijing Academy of Artificial Intelligence / BAAI,Huazhong University of Science and Technology, based in China, categorised as academia,Academia.

When was EVA-CLIP (EVA-02-CLIP-E/14+) released?

EVA-CLIP (EVA-02-CLIP-E/14+) was published in March 2023. 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 EVA-CLIP (EVA-02-CLIP-E/14+) used for?

EVA-CLIP (EVA-02-CLIP-E/14+) works in Vision, and is recorded as handling image classification. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Source

Original publication

Record last updated 11 February 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