SambaLingo-Thai-Chat (7B) TPS calculator

Open weights SambaNova Systems, Inc 7B parameters April 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

589 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Tesla K20c

5 GB · IQ4_XS · 26.5 tok/s

Fastest card

B200

488 tok/s · 180 GB

Which GPUs can run SambaLingo-Thai-Chat (7B)?

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.

589 cards match

Calculating
Needs Quantisation Fit
488 tok/s

293–780 · low confidence

B200 NVIDIA 180 GB 8,000 GB/s Jan 2024 8.1 GB Q8_0 Comfortable
488 tok/s

293–780 · low confidence

B300 NVIDIA 288 GB 8,000 GB/s Sep 2025 8.1 GB Q8_0 Comfortable
389 tok/s

234–623 · low confidence

Radeon Instinct MI350X AMD 288 GB 8,190 GB/s Jan 2025 8.1 GB Q8_0 Comfortable
389 tok/s

234–623 · low confidence

Radeon Instinct MI355X AMD 288 GB 8,190 GB/s Jan 2025 8.1 GB Q8_0 Comfortable
311 tok/s

187–498 · low confidence

Radeon Instinct MI300 AMD 128 GB 6,550 GB/s Jan 2023 8.1 GB Q8_0 Comfortable
298 tok/s

179–477 · low confidence

H200 NVL NVIDIA 141 GB 4,890 GB/s Nov 2024 8.1 GB Q8_0 Comfortable
298 tok/s

179–477 · low confidence

H200 SXM 141 GB NVIDIA 141 GB 4,890 GB/s Nov 2024 8.1 GB Q8_0 Comfortable
285 tok/s

171–456 · low confidence

Radeon Instinct MI325X AMD 256 GB 6,000 GB/s Oct 2024 8.1 GB Q8_0 Comfortable
253 tok/s

152–405 · low confidence

Radeon Instinct MI300A AMD 128 GB 5,325 GB/s Dec 2023 8.1 GB Q8_0 Comfortable
253 tok/s

152–405 · low confidence

Radeon Instinct MI300X AMD 192 GB 5,325 GB/s Dec 2023 8.1 GB Q8_0 Comfortable
253 tok/s

152–405 · low confidence

Radeon Instinct MI308X AMD 192 GB 5,325 GB/s Dec 2023 8.1 GB Q8_0 Comfortable
240 tok/s

144–384 · low confidence

H100 NVL 94 GB NVIDIA 94 GB 3,940 GB/s Mar 2023 8.1 GB Q8_0 Comfortable
205 tok/s

123–328 · low confidence

H100 PCIe 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 8.1 GB Q8_0 Comfortable
205 tok/s

123–328 · low confidence

H100 SXM5 80 GB NVIDIA 80 GB 3,360 GB/s Oct 2022 8.1 GB Q8_0 Comfortable
205 tok/s

123–328 · low confidence

H100 SXM5 94 GB NVIDIA 94 GB 3,360 GB/s Mar 2023 8.1 GB Q8_0 Comfortable
205 tok/s

123–328 · low confidence

H100 SXM5 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 8.1 GB Q8_0 Comfortable
205 tok/s

123–328 · low confidence

H800 SXM5 NVIDIA 80 GB 3,360 GB/s Mar 2023 8.1 GB Q8_0 Comfortable
156 tok/s

94–249 · low confidence

Radeon Instinct MI250 AMD 128 GB 3,280 GB/s Nov 2021 8.1 GB Q8_0 Comfortable
156 tok/s

94–249 · low confidence

Radeon Instinct MI250X AMD 128 GB 3,280 GB/s Nov 2021 8.1 GB Q8_0 Comfortable
132 tok/s

79–211 · low confidence

CMP 170HX 8 GB NVIDIA 8 GB 1,490 GB/s Sep 2021 6.5 GB Q6_K Tight
130 tok/s

78–208 · low confidence

Data Center GPU Max 1550 Intel 128 GB 3,280 GB/s Jan 2023 8.1 GB Q8_0 Comfortable
127 tok/s

76–203 · low confidence

Data Center GPU Max Subsystem Intel 128 GB 3,210 GB/s Jan 2023 8.1 GB Q8_0 Comfortable
124 tok/s

75–199 · low confidence

A100 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Nov 2020 8.1 GB Q8_0 Comfortable
124 tok/s

75–199 · low confidence

A100X NVIDIA 80 GB 2,040 GB/s Jun 2021 8.1 GB Q8_0 Comfortable
124 tok/s

75–199 · low confidence

A800 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Aug 2022 8.1 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
SambaNova Systems, Inc
Organisation type
Industry
Country
United States of America
Published
8 April 2024
Authors
Zoltan Csaki, Bo Li, Jonathan Li, Qiantong Xu, Pian Pawakapan, Leon Zhang, Yun Du, Hengyu Zhao, Changran Hu, Urmish Thakker

What it does

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

Domain
Language
Task
Chat
Base model
Llama 2-7B

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
7B

https://huggingface.co/sambanovasystems/SambaLingo-Thai-Chat

Training data
tokens
Epochs
4
Batch size
1,024

"Continuous Pre-training: We pack the pretraining mixture into sequences of length 4096 and pretrain with document attention as described in Section 3.2 of Iyer et al. (2022) to ensure we only attend to tokens in the context of the corresponding text document. We train with a global batch size of 1024, sequence length of 4096, maximum learning rate of 1e-4 with cosine decay, warm-up ratio of 0.01 and a weight decay of 0.1. Each expert is trained for a maximum of 4 epochs, following (Muennighoff …

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
8.6 × 10²² FLOP

"Continuous Pre-training: We pack the pretraining mixture into sequences of length 4096 and pretrain with document attention as described in Section 3.2 of Iyer et al. (2022) to ensure we only attend to tokens in the context of the corresponding text document. We train with a global batch size of 1024, sequence length of 4096, maximum learning rate of 1e-4 with cosine decay, warm-up ratio of 0.01 and a weight decay of 0.1. Each expert is trained for a maximum of 4 epochs, following (Muennighoff …

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 (unrestricted)

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
SambaLingo: Teaching Large Language Models New Languages
Last updated
11 February 2026

The extremes

The ten fastest GPUs that run SambaLingo-Thai-Chat (7B)

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 · Q8_0 488 tok/s
  2. 02 B200 180 GB · 8,000 GB/s · Q8_0 488 tok/s
  3. 03 Radeon Instinct MI350X 288 GB · 8,190 GB/s · Q8_0 389 tok/s
  4. 04 Radeon Instinct MI355X 288 GB · 8,190 GB/s · Q8_0 389 tok/s
  5. 05 Radeon Instinct MI300 128 GB · 6,550 GB/s · Q8_0 311 tok/s
  6. 06 H200 NVL 141 GB · 4,890 GB/s · Q8_0 298 tok/s
  7. 07 H200 SXM 141 GB 141 GB · 4,890 GB/s · Q8_0 298 tok/s
  8. 08 Radeon Instinct MI325X 256 GB · 6,000 GB/s · Q8_0 285 tok/s
  9. 09 Radeon Instinct MI300A 128 GB · 5,325 GB/s · Q8_0 253 tok/s
  10. 10 Radeon Instinct MI300X 192 GB · 5,325 GB/s · Q8_0 253 tok/s

The smallest GPUs that still run SambaLingo-Thai-Chat (7B)

The cheapest route in, by memory capacity. A tight fit runs the model but leaves nothing spare for a longer conversation.

  1. 01 Quadro P2200 5 GB · needs 4.5 GB · IQ4_XS · tight 25.5 tok/s
  2. 02 P102-100 5 GB · needs 4.5 GB · IQ4_XS · tight 56.0 tok/s
  3. 03 GeForce GTX 1060 5 GB 5 GB · needs 4.5 GB · IQ4_XS · tight 20.4 tok/s
  4. 04 Quadro P2000 5 GB · needs 4.5 GB · IQ4_XS · tight 17.8 tok/s
  5. 05 Tesla K20s 5 GB · needs 4.5 GB · IQ4_XS · tight 26.5 tok/s
  6. 06 Tesla K20m 5 GB · needs 4.5 GB · IQ4_XS · tight 26.5 tok/s
  7. 07 Tesla K20c 5 GB · needs 4.5 GB · IQ4_XS · tight 26.5 tok/s
  8. 08 RTX 1000 Mobile Ada Generation 6 GB · needs 4.9 GB · Q4_K_M · tight 27.0 tok/s
  9. 09 GeForce RTX 3050 6 GB 6 GB · needs 4.9 GB · Q4_K_M · tight 23.6 tok/s
  10. 10 Arc A380M 6 GB · needs 4.9 GB · Q4_K_M · tight 17.0 tok/s

What the numbers mean

Hardware requirements in practice

Minimum card

Tesla K20c

Memory needed

4.5 GB

Fastest

488 tok/s

SambaLingo-Thai-Chat (7B) is small enough at 7B parameters that hardware is rarely the obstacle — 589 of the cards we track can run it, including cards several years old.

The entry point is the Tesla K20c: 5 GB of memory, IQ4_XS compression, roughly 26.5 tokens per second.

A B200 is the fastest we calculate for it: about 488 tokens per second, from 8,000 GB/s of memory bandwidth.

What this model is

SambaLingo-Thai-Chat (7B) was published by SambaNova Systems, Inc, in United States of America, in April 2024. The organisation is categorised as industry.

It works in Language, and is recorded as doing chat.

It builds on Llama 2-7B, which is why it shares that model's general shape and size.

Because its weights were released, nothing about running it depends on a provider staying available — it is yours once downloaded.

What decides the speed

Half the cards that hold it manage more than 26.3 tokens per second, and 559 exceed reading speed outright.

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.

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.

Training and provenance

Training it took roughly 8.6 × 10²² FLOP of computation — a measure of what producing the model cost, not of how fast it answers.

Step by step

How to choose a GPU for SambaLingo-Thai-Chat (7B)

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 SambaLingo-Thai-Chat (7B) — around 4.5 GB at IQ4_XS. Capacity is the gate — a card either holds it or it does not.

  2. 02

    Match the context to your actual use

    Set the context to what you will actually use. The cache grows with the conversation, and it is the usual reason SambaLingo-Thai-Chat (7B) stops fitting a card that seemed fine.

  3. 03

    Choose how far you will compress it

    The quantisation column varies by card, because a bigger card holds a more accurate copy of SambaLingo-Thai-Chat (7B) — IQ4_XS on the smallest card that fits. Set a floor to hold the comparison at one level.

  4. 04

    Sort by speed

    Sort by speed to see how cards rank for SambaLingo-Thai-Chat (7B). It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 488 tok/s.

  5. 05

    Look at the headroom, not just the fit

    A tight fit runs SambaLingo-Thai-Chat (7B) but leaves nothing spare for a longer conversation; comfortable has headroom. If you expect to grow the context, buy for comfortable.

  6. 06

    Check the card from the other side

    Each card page repeats this sweep for every model we hold. It answers what else the hardware is good for, beyond SambaLingo-Thai-Chat (7B).

Answers

SambaLingo-Thai-Chat (7B) — common questions

01

What GPU do I need to run SambaLingo-Thai-Chat (7B)?

The smallest card in our catalogue that holds SambaLingo-Thai-Chat (7B) is the Tesla K20c, with 5 GB of memory. It runs the model at IQ4_XS using about 4.5 GB, and produces roughly 26.5 tokens per second. 589 cards in total can run it.

02

How fast is SambaLingo-Thai-Chat (7B) on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 488 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 559 of the cards that can run SambaLingo-Thai-Chat (7B) clear that.

03

How much VRAM does SambaLingo-Thai-Chat (7B) need?

About 4.5 GB at IQ4_XS 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.

04

Can I run SambaLingo-Thai-Chat (7B) on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q6_K, using about 6.5 GB and generating roughly 132 tokens per second — a tight fit.

05

Can I run SambaLingo-Thai-Chat (7B) on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 8.1 GB and generating roughly 55.6 tokens per second — a comfortable fit.

06

Can I run SambaLingo-Thai-Chat (7B) on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 8.1 GB and generating roughly 68.9 tokens per second — a comfortable fit.

07

Can I run SambaLingo-Thai-Chat (7B) on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 8.1 GB and generating roughly 81.7 tokens per second — a comfortable fit.

08

Is SambaLingo-Thai-Chat (7B) open source?

Its weights are published, so SambaLingo-Thai-Chat (7B) 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.

09

How many parameters does SambaLingo-Thai-Chat (7B) have?

SambaLingo-Thai-Chat (7B) has 7B parameters. https://huggingface.co/sambanovasystems/SambaLingo-Thai-Chat. 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.

10

Who created SambaLingo-Thai-Chat (7B)?

SambaLingo-Thai-Chat (7B) was published by SambaNova Systems, Inc, based in United States of America, categorised as industry.

11

When was SambaLingo-Thai-Chat (7B) released?

SambaLingo-Thai-Chat (7B) was published in April 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.

12

What is SambaLingo-Thai-Chat (7B) used for?

SambaLingo-Thai-Chat (7B) works in Language, and is recorded as handling chat. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

13

Where can I download SambaLingo-Thai-Chat (7B)?

The weights for SambaLingo-Thai-Chat (7B) are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.

14

How much compute was used to train SambaLingo-Thai-Chat (7B)?

Around 8.6 × 10²² FLOP. 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.

15

Can I run SambaLingo-Thai-Chat (7B) 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 SambaLingo-Thai-Chat (7B) is rarely worth using — the nearest miss we calculate is short by 1.3 GB. Every figure here assumes the whole model is on the card.

16

Would two GPUs run SambaLingo-Thai-Chat (7B) faster?

A second card roughly doubles the memory available but not the generation rate. With 589 cards already able to run SambaLingo-Thai-Chat (7B) alone, the case for pairing is weak.

17

Why does the quantisation differ between cards for SambaLingo-Thai-Chat (7B)?

Each card is shown running the least-compressed copy it can hold, and SambaLingo-Thai-Chat (7B) appears at 4 different compression levels across the cards that fit it. Bigger cards get the more accurate version.

18

How accurate are these SambaLingo-Thai-Chat (7B) 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 293–780 tok/s on the B200 rather than a single number.

Source

Original publication

Record last updated 11 February 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