Llama 3.1-405B TPS calculator

Open weights Meta AI 405B 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

4 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Radeon Instinct MI325X

256 GB · Q4_K_M · 11.3 tok/s

Fastest card

B300

19.3 tok/s · 288 GB

Which GPUs can run Llama 3.1-405B?

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.

4 cards match

Calculating
Needs Quantisation Fit
19.3 tok/s

16–23

B300 NVIDIA 288 GB 8,000 GB/s Sep 2025 224.3 GB Q4_K_M Tight
15.4 tok/s

9–25 · low confidence

Radeon Instinct MI350X AMD 288 GB 8,190 GB/s Jan 2025 224.3 GB Q4_K_M Tight
15.4 tok/s

9–25 · low confidence

Radeon Instinct MI355X AMD 288 GB 8,190 GB/s Jan 2025 224.3 GB Q4_K_M Tight
11.3 tok/s

7–18 · low confidence

Radeon Instinct MI325X AMD 256 GB 6,000 GB/s Oct 2024 224.3 GB Q4_K_M 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
Meta AI
Organisation type
Industry
Country
United States of America
Published
23 July 2024
Authors
Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Alan Schelten, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, Arun Rao, Aston Zhang, Aurelien Rodriguez, Austen Gregerson, Ava Spataru, Baptiste Roziere, Bethany Biron, Binh Tang, Bobbie Chern, Charlotte Caucheteux, Ch…

What it does

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

Domain
Language
Task
Language modeling/generation, Question answering, Code generation, Mathematical reasoning
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
405B

405B

Training data
15,600,000,000,000 tokens

15.6T tokens

Epochs
1
Batch size
16,000,000

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
3.8 × 10²⁵ FLOP

Stated in paper. Also, 6 * 405B * 15.6T training tokens = 3.8e25

How it was established
Reported,Operation counting
Post-training compute
9.4 × 10²² FLOP

Section 4 gives detail about the post-training process. They do 6 rounds of post-training, using the model from the previous iteration in each successive round. In each round, they fine-tune a copy of the language model into a reward model (RM) using preference data, then use the reward model to do rejection sampling on human annotation prompts. Next they do supervised fine-tuning (SFT) on the rejection sampled data along with some synthetic data (8.5k to 9k steps per round). Next, they do Direc…

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 H100 SXM5 80GB
Chips used
16,384
Wall-clock time
2,142 hours (89.3 days)

Trained on 30.84M GPU hours (https://huggingface.co/blog/llama31) and used "up to 16K H100 GPU[s]" so training took at least 30.84M / 16k = 1927.5 hours or ~80 days. Section 3.3.4 gives reliability details over a 54 day period during training, for which they had "higher than 90% effective training time" 1927.5 / 0.9 = 2142 hours Probably, full training time is somewhat longer, since it sounds like there were periods where not all 16k H100s were running.

Hardware utilisation
MFU 40.4%

MFU ranges between 0.38 and 0.43 depending on the specific parallelism used; I assume the geometric mean: sqrt(0.38 * 0.43) = 0.4042

Power draw
22.6 MW
Compute cost
$52,885,434

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 (restricted use)

Llama 3.1 model license: https://huggingface.co/meta-llama/Meta-Llama-3.1-8B/blob/main/LICENSE must seek separate license if over 700m monthly users, acceptable use restrictions training code here: https://github.com/meta-llama/llama-recipes/blob/main/src/llama_recipes/utils/train_utils.py#L70

Hugging Face
meta-llama

How it is classified

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

Frontier model
Yes
Likely above 10²³ FLOP
Yes
Why it is tracked
SOTA improvement,Training cost

High training compute, exceeds 4o and Claude 3.5 on some benchmarks: https://ai.meta.com/blog/meta-llama-3-1/

Record confidence
Confident

Sources

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

Reference
The Llama 3 Herd of Models
Last updated
28 November 2025

The extremes

The ten fastest GPUs that run Llama 3.1-405B

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.

  1. 01 B300 288 GB · 8,000 GB/s · Q4_K_M 19.3 tok/s
  2. 02 Radeon Instinct MI350X 288 GB · 8,190 GB/s · Q4_K_M 15.4 tok/s
  3. 03 Radeon Instinct MI355X 288 GB · 8,190 GB/s · Q4_K_M 15.4 tok/s
  4. 04 Radeon Instinct MI325X 256 GB · 6,000 GB/s · Q4_K_M 11.3 tok/s

What the numbers mean

The hardware side

Minimum card

Radeon Instinct MI325X

Memory needed

224.3 GB

Fastest

19.3 tok/s

At 405B parameters, Llama 3.1-405B is beyond what any single graphics card holds. Running it means either splitting it across several cards or renting hardware built for the job — 4 of the cards we track can hold it on their own, and all of them are datacentre parts.

At the low end, a Radeon Instinct MI325X handles it — 256 GB, at Q4_K_M, for about 11.3 tokens per second.

At the other end, a B300 generates roughly 19.3 tokens per second on it, on the strength of 8,000 GB/s of memory bandwidth.

About this model

Llama 3.1-405B was published by Meta AI, in United States of America, in July 2024. The organisation is categorised as industry.

It works in Language, and is recorded as doing language modeling/generation, Question answering, Code generation, Mathematical reasoning.

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. It is published under the meta-llama organisation on Hugging Face.

How fast it runs, and why

The median result is around 15.4 tokens per second; 4 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 attention layout is on file, so the memory figures are computed exactly rather than approximated.

Training and provenance

Training it took roughly 3.8 × 10²⁵ FLOP of computation, on NVIDIA H100 SXM5 80GB — a measure of what producing the model cost, not of how fast it answers.

It was trained on about 15,600,000,000,000 tokens of text.

Its inclusion criterion is sOTA improvement,Training cost.

Step by step

How to choose a GPU for Llama 3.1-405B

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

  1. 01

    Read the memory figure first

    Every card here has been checked against Llama 3.1-405B — around 224.3 GB at Q4_K_M. Capacity is the gate — a card either holds it or it does not.

  2. 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 Llama 3.1-405B can slip off a card that handles short questions easily.

  3. 03

    Set a quality floor

    Compression is what makes Llama 3.1-405B fit smaller cards, at some cost in accuracy — Q4_K_M on the smallest card that fits. A minimum quality removes the ones that go too far.

  4. 04

    Compare tokens per second, not specifications

    Ranking by tokens per second for Llama 3.1-405B follows memory bandwidth, not core counts, which is why the B300 tops it at 19.3 tok/s.

  5. 05

    Read the fit column last

    Tight means Llama 3.1-405B 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.

  6. 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 Llama 3.1-405B is settled.

Answers

Llama 3.1-405B — common questions

01

Would two GPUs run Llama 3.1-405B faster?

A second card roughly doubles the memory available but not the generation rate. With 4 cards already able to run Llama 3.1-405B alone, the case for pairing is weak.

02

Why does the quantisation differ between cards for Llama 3.1-405B?

A larger card holds a more accurate copy. Across the cards that run Llama 3.1-405B, 1 compression levels are used; the floor control above pins it to one.

03

How accurate are these Llama 3.1-405B 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 16–23 tok/s on the B300 rather than a single number.

04

What GPU do I need to run Llama 3.1-405B?

The smallest card in our catalogue that holds Llama 3.1-405B is the Radeon Instinct MI325X, with 256 GB of memory. It runs the model at Q4_K_M using about 224.3 GB, and produces roughly 11.3 tokens per second. 4 cards in total can run it.

05

How fast is Llama 3.1-405B on a GPU?

It depends on the card. The quickest we calculate is a B300 at about 19.3 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 4 of the cards that can run Llama 3.1-405B clear that.

06

How much VRAM does Llama 3.1-405B need?

About 224.3 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.

07

Is Llama 3.1-405B open source?

Its weights are published, so Llama 3.1-405B 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

How many parameters does Llama 3.1-405B have?

Llama 3.1-405B has 405B parameters. 405B. 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

Who created Llama 3.1-405B?

Llama 3.1-405B was published by Meta AI, based in United States of America, categorised as industry.

10

When was Llama 3.1-405B released?

Llama 3.1-405B 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

What is Llama 3.1-405B used for?

Llama 3.1-405B works in Language, and is recorded as handling language modeling/generation, Question answering, Code generation, Mathematical reasoning. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

12

Where can I download Llama 3.1-405B?

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

13

How much compute was used to train Llama 3.1-405B?

Around 3.8 × 10²⁵ FLOP, on NVIDIA H100 SXM5 80GB. 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.

14

Can I run Llama 3.1-405B if it does not fit in my GPU?

It can be split between the card and system memory, but Llama 3.1-405B generates painfully slowly that way — the nearest miss we calculate is short by 51.5 GB. Nothing on this page assumes offloading.

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.