Gemma 2 9B TPS calculator

Open weights Google DeepMind 9B parameters June 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

509 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Xeon Phi 5110P

8 GB · IQ4_XS · 24.0 tok/s

Fastest card

B200

376 tok/s · 180 GB

Which GPUs can run Gemma 2 9B?

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
376 tok/s 320–452	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	11.7 GB	Q8_0	Comfortable
376 tok/s 320–452	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	11.7 GB	Q8_0	Comfortable
301 tok/s 180–481 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	11.7 GB	Q8_0	Comfortable
301 tok/s 180–481 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	11.7 GB	Q8_0	Comfortable
240 tok/s 144–385 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	11.7 GB	Q8_0	Comfortable
230 tok/s 196–276	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	11.7 GB	Q8_0	Comfortable
230 tok/s 196–276	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	11.7 GB	Q8_0	Comfortable
220 tok/s 132–352 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	11.7 GB	Q8_0	Comfortable
195 tok/s 117–313 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	11.7 GB	Q8_0	Comfortable
195 tok/s 117–313 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	11.7 GB	Q8_0	Comfortable
195 tok/s 117–313 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	11.7 GB	Q8_0	Comfortable
185 tok/s 158–222	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	11.7 GB	Q8_0	Comfortable
172 tok/s 146–207	CMP 170HX 8 GB NVIDIA	8 GB	1,490 GB/s	Sep 2021	7.0 GB	IQ4_XS	Tight
158 tok/s 134–190	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	11.7 GB	Q8_0	Comfortable
158 tok/s 134–190	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	11.7 GB	Q8_0	Comfortable
158 tok/s 134–190	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	11.7 GB	Q8_0	Comfortable
158 tok/s 134–190	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	11.7 GB	Q8_0	Comfortable
158 tok/s 134–190	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	11.7 GB	Q8_0	Comfortable
131 tok/s 111–157	CMP 170HX 10 GB NVIDIA	10 GB	1,560 GB/s	Sep 2021	8.6 GB	Q5_K_M	Tight
120 tok/s 72–193 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	11.7 GB	Q8_0	Comfortable
120 tok/s 72–193 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	11.7 GB	Q8_0	Comfortable
100 tok/s 60–161 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	11.7 GB	Q8_0	Comfortable
98.2 tok/s 59–157 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	11.7 GB	Q8_0	Comfortable
96.0 tok/s 82–115	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	11.7 GB	Q8_0	Comfortable
96.0 tok/s 82–115	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	11.7 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: Google DeepMind
Organisation type: Industry
Country: United States of America
Published: 24 June 2024
Authors: Gemma Team, Google DeepMind

What it does

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

Domain: Language
Task: Language modeling/generation, Chat, Code generation, Question answering, Quantitative reasoning

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: 9B
Training data: 8,000,000,000,000 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: 4.3 × 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: Google TPU v4
Chips used: 4,096
Power draw: 2.7 MW

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: Unreleased
Hugging Face: google

How it is classified

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

Likely above 10²³ FLOP: Yes
Record confidence: Confident

Sources

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

Reference: Gemma 2 offers best-in-class performance, runs at incredible speed across different hardware and easily integrates with other AI tools.
Last updated: 28 November 2025

The extremes

The ten fastest GPUs that run Gemma 2 9B

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 Gemma 2 9B

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

What it takes to run this model

Minimum card

Xeon Phi 5110P

Memory needed

7.0 GB

Fastest

376 tok/s

Gemma 2 9B reaches a parameter count of 9B. That is small enough that hardware is rarely the obstacle, including on cards several years old. The number of cards we track that can run it: 509.

The smallest card that holds it is Xeon Phi 5110P, with a memory capacity of 8 GB, running it at a compression of IQ4_XS and producing around 24.0 tokens per second.

The quickest result comes from B200, generating roughly 376 tokens per second on the strength of a memory bandwidth of 8,000 GB/s.

Background

Gemma 2 9B was published by Google DeepMind, in the country recorded as United States of America, during June 2024. 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/generation, Chat, Code generation, Question answering, Quantitative reasoning.

Because its weights were released, nothing about running it depends on a provider staying available — it is yours once downloaded. On Hugging Face it is published under the organisation google.

Reading the throughput figures

Half the cards that hold it manage more than 26.0 tokens per second. Exceeding reading speed outright: 473 of them.

Every weight participates in every token here, so bandwidth is the whole story: the ranking below is effectively a ranking of memory throughput.

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

What went into building it

Producing it required arithmetic totalling around 4.3 × 10²³ FLOP, on hardware recorded as Google TPU v4. That figure measures what producing the model cost, and has no bearing on how fast it answers.

It was trained on a corpus of about 8,000,000,000,000 tokens of text.

Step by step

How to choose a GPU for Gemma 2 9B

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

The table lists every card able to hold Gemma 2 9B, needing around 7.0 GB at a compression of IQ4_XS. That figure, not the headline performance of a card, is what decides whether it runs.
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 Gemma 2 9B.
03

Decide how much compression you will accept

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

Compare tokens per second, not specifications

The speed ordering is effectively an ordering by memory bandwidth, for Gemma 2 9B. It will not match a gaming ordering, because generation is bound by memory bandwidth. The card topping the list is B200, at 376 tok/s.
05

Look at the headroom, not just the fit

The fit column separates cards that just manage it from those with room to spare, in the case of Gemma 2 9B. 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 Gemma 2 9B.

Answers

Gemma 2 9B — common questions

Gemma 2 9B— 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.

Gemma 2 9B— 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: 320–452 tok/s on B200. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

Gemma 2 9B— what GPU do I need to run it?

The smallest card in our catalogue that holds it is Xeon Phi 5110P, with a memory capacity of 8 GB. It runs the model at a compression of IQ4_XS using about 7.0 GB, and produces roughly 24.0 tokens per second. The number of cards able to run it in total: 509.

Gemma 2 9B— how fast is it on a GPU?

It depends on the card. The quickest we calculate is B200, at about 376 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: 473.

Gemma 2 9B— how much VRAM does it need?

It needs about 7.0 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.

Gemma 2 9B— can I run it on a GPU holding 8 GB?

Yes. The card CMP 170HX 8 GB, holding 8 GB, runs it at a compression of IQ4_XS, using about 7.0 GB and generating roughly 172 tokens per second. The fit is tight.

Gemma 2 9B— can I run it on a GPU holding 12 GB?

Yes. The card GeForce RTX 3080 Ti, holding 12 GB, runs it at a compression of Q6_K, using about 9.6 GB and generating roughly 62.4 tokens per second. The fit is tight.

Gemma 2 9B— can I run it on a GPU holding 16 GB?

Yes. The card Tesla V100 SXM2 16 GB, holding 16 GB, runs it at a compression of Q8_0, using about 11.7 GB and generating roughly 53.2 tokens per second. The fit is comfortable.

Gemma 2 9B— can I run it on a GPU holding 24 GB?

Yes. The card GeForce RTX 5090 D V2, holding 24 GB, runs it at a compression of Q8_0, using about 11.7 GB and generating roughly 63.1 tokens per second. The fit is comfortable.

Gemma 2 9B— 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.

Gemma 2 9B— how many parameters does it have?

It has a parameter count of 9B. 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.

Gemma 2 9B— who created it?

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

Gemma 2 9B— when was it released?

It was published in June 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.

Gemma 2 9B— what is it used for?

It works in the domain of Language, and is recorded as handling the task of language modeling/generation, Chat, Code generation, Question answering, Quantitative reasoning. These are the areas it was designed around; they describe intent rather than a hard boundary.

Gemma 2 9B— where can I download it?

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

Gemma 2 9B— how much compute was used to train it?

Training consumed around 4.3 × 10²³ FLOP, on hardware recorded as Google TPU v4. 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.

Gemma 2 9B— 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 2.2 GB. Every figure here assumes the whole model is resident on the card.

Gemma 2 9B— would two GPUs run it faster?

Two cards buy memory rather than speed, which matters only if one card cannot hold it. The number that can: 509. So a second card is rarely the answer here.

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.

All AI models