Shuka-1 TPS calculator

Open weights Sarvam 9.6B parameters August 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

582 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Quadro 6000

6 GB · Q3_K_M · 14.6 tok/s

Fastest card

B200

354 tok/s · 180 GB

Which GPUs can run Shuka-1?

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

582 cards match

Calculating

					Needs	Quantisation	Fit
354 tok/s 213–567 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	10.9 GB	Q8_0	Comfortable
354 tok/s 213–567 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	10.9 GB	Q8_0	Comfortable
283 tok/s 170–453 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	10.9 GB	Q8_0	Comfortable
283 tok/s 170–453 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	10.9 GB	Q8_0	Comfortable
226 tok/s 136–362 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	10.9 GB	Q8_0	Comfortable
217 tok/s 130–347 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	10.9 GB	Q8_0	Comfortable
217 tok/s 130–347 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	10.9 GB	Q8_0	Comfortable
207 tok/s 124–332 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	10.9 GB	Q8_0	Comfortable
184 tok/s 110–294 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	10.9 GB	Q8_0	Comfortable
184 tok/s 110–294 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	10.9 GB	Q8_0	Comfortable
184 tok/s 110–294 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	10.9 GB	Q8_0	Comfortable
175 tok/s 105–279 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	10.9 GB	Q8_0	Comfortable
152 tok/s 91–244 · low confidence	CMP 170HX 8 GB NVIDIA	8 GB	1,490 GB/s	Sep 2021	6.5 GB	Q4_K_M	Tight
149 tok/s 89–238 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	10.9 GB	Q8_0	Comfortable
149 tok/s 89–238 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	10.9 GB	Q8_0	Comfortable
149 tok/s 89–238 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	10.9 GB	Q8_0	Comfortable
149 tok/s 89–238 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	10.9 GB	Q8_0	Comfortable
149 tok/s 89–238 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	10.9 GB	Q8_0	Comfortable
113 tok/s 68–181 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	10.9 GB	Q8_0	Comfortable
113 tok/s 68–181 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	10.9 GB	Q8_0	Comfortable
100 tok/s 60–161 · low confidence	CMP 170HX 10 GB NVIDIA	10 GB	1,560 GB/s	Sep 2021	8.7 GB	Q6_K	Tight
94.5 tok/s 57–151 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	10.9 GB	Q8_0	Comfortable
92.4 tok/s 55–148 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	10.9 GB	Q8_0	Comfortable
90.4 tok/s 54–145 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	10.9 GB	Q8_0	Comfortable
90.4 tok/s 54–145 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	10.9 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: Sarvam
Organisation type: Industry
Country: India
Published: 30 August 2024

What it does

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

Domain: Speech
Task: Speech recognition (ASR), Speech-to-text
Base model: Saaras v1,Llama 3-8B

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: 9.6B
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: Unreleased
Hugging Face: sarvamai

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: Shuka v1: Revolutionizing Audio Understanding for Indic Languages
Last updated: 28 November 2025

The extremes

The ten fastest GPUs that run Shuka-1

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 Shuka-1

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

Quadro 6000

Memory needed

5.4 GB

Fastest

354 tok/s

Shuka-1 reaches a parameter count of 9.6B. 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: 582.

At the low end it is handled by Quadro 6000, with a memory capacity of 6 GB, running it at a compression of Q3_K_M and producing around 14.6 tokens per second.

At the other end sits B200, generating roughly 354 tokens per second on the strength of a memory bandwidth of 8,000 GB/s.

Background

Shuka-1 was published by Sarvam, in the country recorded as India, during August 2024. The category the publisher falls under is industry.

It works in the domain of Speech, and is recorded as performing the task of speech recognition (ASR), Speech-to-text.

Its starting point was an existing base model, Saaras v1,Llama 3-8B. That is the usual way a specialised model is produced.

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 sarvamai.

Reading the throughput figures

Half the cards that hold it manage more than 23.0 tokens per second. Clearing the ten tokens per second that roughly matches reading speed: 545 of them.

It is a dense model, so every parameter is read for every token produced. That makes speed track memory bandwidth almost exactly — a card with twice the bandwidth generates roughly twice as fast.

Memory here is estimated from size rather than computed from the architecture, which is not recorded for this model — the numbers are indicative rather than exact.

Step by step

How to choose a GPU for Shuka-1

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

01

Start from the memory column

The table lists every card able to hold Shuka-1, needing around 5.4 GB at a compression of Q3_K_M. Capacity is the gate — a card either holds it or it does not.
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 Shuka-1.
03

Choose how far you will compress it

Compression is what makes a model fit smaller cards, at some cost in accuracy, 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

Rank by throughput rather than spec sheet

Ranking by tokens per second follows memory bandwidth rather than core counts, for Shuka-1. It will not match a gaming ordering, because generation is bound by memory bandwidth. The card topping the list is B200, at 354 tok/s.
05

Read the fit column last

A tight fit runs, but leaves nothing spare for a longer conversation, in the case of Shuka-1. 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

Following a card through to its own page shows every other model it can hold, which is the question that follows once you have settled on Shuka-1.

Answers

Shuka-1 — common questions

Shuka-1— how many parameters does it have?

It has a parameter count of 9.6B. "Our state-of-the-art, in-house, audio encoder: Saaras v1 Meta’s Llama3-8B-Instruct as the decoder The encoder and decoder are connected by a small projector with ~60M parameters.". 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.

Shuka-1— who created it?

It was published by Sarvam, based in India, an organisation categorised as industry.

Shuka-1— when was it released?

It was published in August 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.

Shuka-1— what is it used for?

It works in the domain of Speech, and is recorded as handling the task of speech recognition (ASR), Speech-to-text. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Shuka-1— where can I download it?

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

Shuka-1— can I run it if it does not fit in my GPU?

Only by offloading, which is usually a false economy: the part held in system memory drags the whole thing down. The nearest miss we calculate falls short by 2.0 GB. Every figure here assumes the whole model is resident on the card.

Shuka-1— would two GPUs run it faster?

Capacity adds across cards; throughput does not. The number of cards already holding it on their own: 582. So a second card is rarely the answer here.

Shuka-1— 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.

Shuka-1— how accurate are these speed estimates?

They are calculated from specifications rather than measured, and each carries a range. One example: 213–567 tok/s on B200. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

Shuka-1— what GPU do I need to run it?

The smallest card in our catalogue that holds it is Quadro 6000, with a memory capacity of 6 GB. It runs the model at a compression of Q3_K_M using about 5.4 GB, and produces roughly 14.6 tokens per second. The number of cards able to run it in total: 582.

Shuka-1— how fast is it on a GPU?

It depends on the card. The quickest we calculate is B200, at about 354 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: 545.

Shuka-1— how much VRAM does it need?

It needs about 5.4 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.

Shuka-1— 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 Q4_K_M, using about 6.5 GB and generating roughly 152 tokens per second. The fit is tight.

Shuka-1— 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 8.7 GB and generating roughly 58.7 tokens per second. The fit is comfortable.

Shuka-1— 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 10.9 GB and generating roughly 50.1 tokens per second. The fit is comfortable.

Shuka-1— 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 10.9 GB and generating roughly 59.4 tokens per second. The fit is comfortable.

Shuka-1— 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.

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