LLaMA-65B TPS calculator

Open weights Meta AI 65.2B parameters February 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

61 cards that can run it

818 cards we hold specifications for

Smallest card that fits

A100 PCIe 40 GB

40 GB · Q3_K_M · 27.3 tok/s

Fastest card

B200

52.0 tok/s · 180 GB

Which GPUs can run LLaMA-65B?

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

61 cards match

Calculating

					Needs	Quantisation	Fit
52.0 tok/s 44–62	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	71.4 GB	Q8_0	Comfortable
52.0 tok/s 44–62	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	71.4 GB	Q8_0	Comfortable
41.5 tok/s 25–66 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	71.4 GB	Q8_0	Comfortable
41.5 tok/s 25–66 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	71.4 GB	Q8_0	Comfortable
33.2 tok/s 20–53 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	71.4 GB	Q8_0	Comfortable
31.8 tok/s 27–38	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	71.4 GB	Q8_0	Comfortable
31.8 tok/s 27–38	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	71.4 GB	Q8_0	Comfortable
30.4 tok/s 18–49 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	71.4 GB	Q8_0	Comfortable
28.0 tok/s 24–34	GRID A100B NVIDIA	48 GB	1,870 GB/s	May 2020	41.1 GB	Q4_K_M	Tight
27.3 tok/s 23–33	A100 PCIe 40 GB NVIDIA	40 GB	1,560 GB/s	Jun 2020	33.5 GB	Q3_K_M	Tight
27.3 tok/s 23–33	A100 SXM4 40 GB NVIDIA	40 GB	1,560 GB/s	May 2020	33.5 GB	Q3_K_M	Tight
27.3 tok/s 23–33	A800 PCIe 40 GB NVIDIA	40 GB	1,560 GB/s	Nov 2022	33.5 GB	Q3_K_M	Tight
27.0 tok/s 16–43 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	71.4 GB	Q8_0	Comfortable
27.0 tok/s 16–43 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	71.4 GB	Q8_0	Comfortable
27.0 tok/s 16–43 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	71.4 GB	Q8_0	Comfortable
25.6 tok/s 22–31	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	71.4 GB	Q8_0	Tight
21.8 tok/s 19–26	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	71.4 GB	Q8_0	Comfortable
21.8 tok/s 19–26	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	71.4 GB	Q8_0	Tight
21.8 tok/s 19–26	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	71.4 GB	Q8_0	Tight
21.8 tok/s 19–26	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	71.4 GB	Q8_0	Comfortable
21.8 tok/s 19–26	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	71.4 GB	Q8_0	Tight
20.1 tok/s 17–24	RTX PRO 5000 Blackwell NVIDIA	48 GB	1,340 GB/s	Mar 2025	41.1 GB	Q4_K_M	Tight
19.1 tok/s 16–23	H100 SXM5 64 GB NVIDIA	64 GB	2,020 GB/s	Mar 2023	56.3 GB	Q6_K	Tight
16.6 tok/s 10–27 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	71.4 GB	Q8_0	Comfortable
16.6 tok/s 10–27 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	71.4 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: Meta AI
Organisation type: Industry
Country: United States of America
Published: 24 February 2023
Authors: Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, Guillaume Lample

What it does

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

Domain: Language
Task: Language modeling, Code generation
Approach: Supervised

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: 65.2B
Training data: 1,400,000,000,000 tokens
Epochs: 1.09
Batch size: 4,000,000

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: 5.5 × 10²³ 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 A100
Chips used: 2,048
Chip-hours: 1,024,000
Wall-clock time: 500 hours (20.8 days)
Hardware utilisation: MFU 47.5%
Power draw: 1.6 MW
Compute cost: $578,026

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 (non-commercial)
Training code: Unreleased

How it is classified

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

Foundation model: Yes
Likely above 10²³ FLOP: Yes
Why it is tracked: Historical significance,Highly cited
Record confidence: Confident
Citations: 19,926
Benchmark data: LLaMA-65B

Sources

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

Reference: LLaMA: Open and Efficient Foundation Language Models
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run LLaMA-65B

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 LLaMA-65B

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

A100 PCIe 40 GB

Memory needed

33.5 GB

Fastest

52.0 tok/s

LLaMA-65B reaches a parameter count of 65.2B. 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: 61.

The entry point is A100 PCIe 40 GB, with a memory capacity of 40 GB, running it at a compression of Q3_K_M and producing around 27.3 tokens per second.

The quickest result comes from B200, generating roughly 52.0 tokens per second on the strength of a memory bandwidth of 8,000 GB/s.

What this model is

LLaMA-65B was published by Meta AI, in the country recorded as United States of America, during February 2023. It comes out of an organisation categorised as industry.

It works in the domain of Language, and is recorded as performing the task of language modeling, Code generation.

The weights are published, so it can be downloaded and run on your own hardware indefinitely, offline, with no account attached.

What decides the speed

The median result is around 13.3 tokens per second. Producing text faster than most people read it: 56 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.

Because the architecture is recorded, the memory column is derived rather than estimated.

What went into building it

Training it took a computation budget of roughly 5.5 × 10²³ FLOP, on hardware recorded as NVIDIA A100. That figure measures what producing the model cost, and has no bearing on how fast it answers.

Training consumed a corpus of around 1,400,000,000,000 tokens of text.

Its inclusion criterion: historical significance,Highly cited.

Step by step

How to choose a GPU for LLaMA-65B

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

Every card here has been checked against LLaMA-65B, needing around 33.5 GB at a compression of Q3_K_M. Capacity is the gate — a card either holds it or it does not.
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, because at long context a card that handles short questions easily can be dropped by LLaMA-65B.
03

Choose how far you will compress it

The quantisation column varies by card, because a bigger card holds a more accurate copy, reaching a compression of Q3_K_M 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.
04

Sort by speed

Ranking by tokens per second follows memory bandwidth rather than core counts, for LLaMA-65B. It will not match a gaming ordering, because generation is bound by memory bandwidth. The card topping the list is B200, at 52.0 tok/s.
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 LLaMA-65B. 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.
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. A card is usually bought for more than one model, so it is worth a look before buying for LLaMA-65B.

Answers

LLaMA-65B — common questions

LLaMA-65B— would two GPUs run it faster?

Capacity adds across cards; throughput does not. The number of cards already holding it on their own: 61. So a second card is rarely the answer here.

LLaMA-65B— why does the quantisation differ between cards?

Because capacity varies, so does how hard it has to be squeezed. The number of distinct levels in the table above: 4. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.

LLaMA-65B— 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: 44–62 tok/s on B200. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

LLaMA-65B— what GPU do I need to run it?

The smallest card in our catalogue that holds it is A100 PCIe 40 GB, with a memory capacity of 40 GB. It runs the model at a compression of Q3_K_M using about 33.5 GB, and produces roughly 27.3 tokens per second. The number of cards able to run it in total: 61.

LLaMA-65B— how fast is it on a GPU?

It depends on the card. The quickest we calculate is B200, at about 52.0 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: 56.

LLaMA-65B— how much VRAM does it need?

It needs about 33.5 GB at a compression of Q3_K_M, 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.

LLaMA-65B— 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.

LLaMA-65B— how many parameters does it have?

It has a parameter count of 65.2B. Model card, table 1: https://github.com/facebookresearch/llama/blob/53011c3d7946dadb8274a4c5c7586ab54edf792d/MODEL_CARD.md. 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.

LLaMA-65B— who created it?

It was published by Meta AI, based in United States of America, an organisation categorised as industry.

LLaMA-65B— when was it released?

It was published in February 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.

LLaMA-65B— what is it used for?

It works in the domain of Language, and is recorded as handling the task of language modeling, Code generation. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.

LLaMA-65B— where can I download it?

The weights are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.

LLaMA-65B— how much compute was used to train it?

Training consumed around 5.5 × 10²³ FLOP, on hardware recorded as NVIDIA A100. 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.

LLaMA-65B— can I run it if it does not fit in my GPU?

It can be split between the card and system memory, but it generates painfully slowly that way. The nearest miss we calculate falls short by 12.3 GB. Every figure here assumes the whole model is resident on the card.

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