InternVL2-Llama3-76B TPS calculator

Open weights Shanghai AI Lab 76B parameters July 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

58 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Quadro RTX 8000

48 GB · IQ4_XS · 9.2 tok/s

Fastest card

B200

44.6 tok/s · 180 GB

Which GPUs can run InternVL2-Llama3-76B?

Set the inputs, read the answer

A longer conversation needs more memory, which can push this model off smaller cards.

Hides cards that would only fit the model by compressing it below this point.

58 cards match

Calculating
Needs Quantisation Fit
44.6 tok/s

27–71 · low confidence

B200 NVIDIA 180 GB 8,000 GB/s Jan 2024 82.1 GB Q8_0 Comfortable
44.6 tok/s

27–71 · low confidence

B300 NVIDIA 288 GB 8,000 GB/s Sep 2025 82.1 GB Q8_0 Comfortable
35.6 tok/s

21–57 · low confidence

Radeon Instinct MI350X AMD 288 GB 8,190 GB/s Jan 2025 82.1 GB Q8_0 Comfortable
35.6 tok/s

21–57 · low confidence

Radeon Instinct MI355X AMD 288 GB 8,190 GB/s Jan 2025 82.1 GB Q8_0 Comfortable
28.5 tok/s

17–46 · low confidence

Radeon Instinct MI300 AMD 128 GB 6,550 GB/s Jan 2023 82.1 GB Q8_0 Comfortable
27.3 tok/s

16–44 · low confidence

H200 NVL NVIDIA 141 GB 4,890 GB/s Nov 2024 82.1 GB Q8_0 Comfortable
27.3 tok/s

16–44 · low confidence

H200 SXM 141 GB NVIDIA 141 GB 4,890 GB/s Nov 2024 82.1 GB Q8_0 Comfortable
27.2 tok/s

16–44 · low confidence

H100 SXM5 80 GB NVIDIA 80 GB 3,360 GB/s Oct 2022 64.4 GB Q6_K Tight
27.2 tok/s

16–44 · low confidence

H800 SXM5 NVIDIA 80 GB 3,360 GB/s Mar 2023 64.4 GB Q6_K Tight
26.1 tok/s

16–42 · low confidence

Radeon Instinct MI325X AMD 256 GB 6,000 GB/s Oct 2024 82.1 GB Q8_0 Comfortable
25.6 tok/s

15–41 · low confidence

GRID A100B NVIDIA 48 GB 1,870 GB/s May 2020 42.3 GB IQ4_XS Tight
23.2 tok/s

14–37 · low confidence

Radeon Instinct MI300A AMD 128 GB 5,325 GB/s Dec 2023 82.1 GB Q8_0 Comfortable
23.2 tok/s

14–37 · low confidence

Radeon Instinct MI300X AMD 192 GB 5,325 GB/s Dec 2023 82.1 GB Q8_0 Comfortable
23.2 tok/s

14–37 · low confidence

Radeon Instinct MI308X AMD 192 GB 5,325 GB/s Dec 2023 82.1 GB Q8_0 Comfortable
22.0 tok/s

13–35 · low confidence

H100 NVL 94 GB NVIDIA 94 GB 3,940 GB/s Mar 2023 82.1 GB Q8_0 Tight
20.1 tok/s

12–32 · low confidence

H100 SXM5 64 GB NVIDIA 64 GB 2,020 GB/s Mar 2023 55.5 GB Q5_K_M Tight
18.7 tok/s

11–30 · low confidence

H100 PCIe 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 82.1 GB Q8_0 Tight
18.7 tok/s

11–30 · low confidence

H100 SXM5 94 GB NVIDIA 94 GB 3,360 GB/s Mar 2023 82.1 GB Q8_0 Tight
18.7 tok/s

11–30 · low confidence

H100 SXM5 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 82.1 GB Q8_0 Tight
18.3 tok/s

11–29 · low confidence

RTX PRO 5000 Blackwell NVIDIA 48 GB 1,340 GB/s Mar 2025 42.3 GB IQ4_XS Tight
16.5 tok/s

10–26 · low confidence

A100 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Nov 2020 64.4 GB Q6_K Tight
16.5 tok/s

10–26 · low confidence

A100X NVIDIA 80 GB 2,040 GB/s Jun 2021 64.4 GB Q6_K Tight
16.5 tok/s

10–26 · low confidence

A800 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Aug 2022 64.4 GB Q6_K Tight
16.5 tok/s

10–26 · low confidence

H100 CNX NVIDIA 80 GB 2,040 GB/s Mar 2023 64.4 GB Q6_K Tight
16.5 tok/s

10–26 · low confidence

H100 PCIe 80 GB NVIDIA 80 GB 2,040 GB/s Oct 2022 64.4 GB Q6_K Tight

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
Shanghai AI Lab
Organisation type
Academia
Country
China
Published
15 July 2024

What it does

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

Domain
Multimodal, Language, Vision, Video
Task
Visual question answering, Language modeling/generation, Question answering, Video description
Base model
Hermes 2 Theta Llama-3 70B,InternViT-6B

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
76B
Training data
tokens

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 (restricted use)
Training code
Open source

llama3 license https://huggingface.co/OpenGVLab/InternVL2-Llama3-76B "This project is released under the MIT License. This project uses the pre-trained Hermes-2-Theta-Llama-3-70B as a component, which is licensed under the Llama 3 Community License." MIT license https://github.com/OpenGVLab/InternVL "Release training / evaluation code for InternVL2 series" is checked off

Hugging Face
OpenGVLab

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
OpenGVLab/InternVL2-Llama3-76B
Last updated
28 November 2025

The extremes

What the numbers mean

What you need to run it

Minimum card

Quadro RTX 8000

Memory needed

42.3 GB

Fastest

44.6 tok/s

InternVL2-Llama3-76B reaches a parameter count of 76B. 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: 58.

The least hardware that works is Quadro RTX 8000, with a memory capacity of 48 GB, running it at a compression of IQ4_XS and producing around 9.2 tokens per second.

The quickest result comes from B200, generating roughly 44.6 tokens per second on the strength of a memory bandwidth of 8,000 GB/s.

What this model is

InternVL2-Llama3-76B was published by Shanghai AI Lab, in the country recorded as China, during July 2024. It comes out of an organisation categorised as academia.

It works in the domain of Multimodal, Language, Vision, Video, and is recorded as performing the task of visual question answering, Language modeling/generation, Question answering, Video description.

Rather than being trained from scratch, it is derived from Hermes 2 Theta Llama-3 70B,InternViT-6B. That is why it shares the base model's general shape and size.

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 OpenGVLab.

What decides the speed

The median result is around 14.3 tokens per second. Exceeding reading speed outright: 43 of them.

Being dense, it reads all of itself per token, which is why the ordering by speed below follows the ordering by memory bandwidth so closely.

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.

Step by step

How to choose a GPU for InternVL2-Llama3-76B

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

  1. 01

    Start from the memory column

    The table lists every card able to hold InternVL2-Llama3-76B, needing around 42.3 GB at a compression of IQ4_XS. That figure, not the headline performance of a card, is what decides whether it runs.

  2. 02

    Decide how long your conversations run

    The conversation occupies memory too, and grows as it goes. Set the slider to the length you expect, because at long context a card that handles short questions easily can be dropped by InternVL2-Llama3-76B.

  3. 03

    Choose how far you will compress it

    Each card runs the least-compressed copy it can hold, reaching a compression of IQ4_XS 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.

  4. 04

    Compare tokens per second, not specifications

    Sort by speed to see how cards rank for InternVL2-Llama3-76B. It will not match a gaming ordering, because generation is bound by memory bandwidth. The card topping the list is B200, at 44.6 tok/s.

  5. 05

    Look at the headroom, not just the fit

    A tight fit runs, but leaves nothing spare for a longer conversation, in the case of InternVL2-Llama3-76B. 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.

  6. 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. A card is usually bought for more than one model, so it is worth a look before buying for InternVL2-Llama3-76B.

Answers

InternVL2-Llama3-76B — common questions

01

InternVL2-Llama3-76B— would two GPUs run it faster?

A second card roughly doubles the memory available but not the generation rate. The number already able to run it alone: 58. So a second card is rarely the answer here.

02

InternVL2-Llama3-76B— why does the quantisation differ between cards?

Each card is shown running the least-compressed copy it can hold. The number of distinct compression levels across the cards that fit it: 4. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.

03

InternVL2-Llama3-76B— how accurate are these speed estimates?

Every figure is derived from memory bandwidth and model size, not benchmarked, which is why each is published as a range rather than a single number. One example: 27–71 tok/s on B200. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

04

InternVL2-Llama3-76B— what GPU do I need to run it?

The smallest card in our catalogue that holds it is Quadro RTX 8000, with a memory capacity of 48 GB. It runs the model at a compression of IQ4_XS using about 42.3 GB, and produces roughly 9.2 tokens per second. The number of cards able to run it in total: 58.

05

InternVL2-Llama3-76B— how fast is it on a GPU?

It depends on the card. The quickest we calculate is B200, at about 44.6 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: 43.

06

InternVL2-Llama3-76B— how much VRAM does it need?

It needs about 42.3 GB at a compression of IQ4_XS, 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.

07

InternVL2-Llama3-76B— 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.

08

InternVL2-Llama3-76B— how many parameters does it have?

It has a parameter count of 76B. 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.

09

InternVL2-Llama3-76B— who created it?

It was published by Shanghai AI Lab, based in China, an organisation categorised as academia.

10

InternVL2-Llama3-76B— when was it released?

It was published in July 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.

11

InternVL2-Llama3-76B— what is it used for?

It works in the domain of Multimodal, Language, Vision, Video, and is recorded as handling the task of visual question answering, Language modeling/generation, Question answering, Video description. These are the areas it was designed around; they describe intent rather than a hard boundary.

12

InternVL2-Llama3-76B— where can I download it?

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

13

InternVL2-Llama3-76B— can I run it 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 model is rarely worth using. The nearest miss we calculate falls short by 10.7 GB. Every figure here assumes the whole model is resident on the card.

Source

Original publication

Record last updated 28 November 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.