XVERSE-65B-2 TPS calculator
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
818 cards we hold specifications for
Smallest card that fits
A100 PCIe 40 GB
40 GB · Q3_K_M · 27.4 tok/s
Fastest card
B200
52.1 tok/s · 180 GB
Which GPUs can run XVERSE-65B-2?
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.
61 cards match
Calculating| Needs | Quantisation | Fit | |||||
|---|---|---|---|---|---|---|---|
|
52.1
tok/s
31–83 · low confidence |
B200 NVIDIA | 180 GB | 8,000 GB/s | Jan 2024 | 70.3 GB | Q8_0 | Comfortable |
|
52.1
tok/s
31–83 · low confidence |
B300 NVIDIA | 288 GB | 8,000 GB/s | Sep 2025 | 70.3 GB | Q8_0 | Comfortable |
|
41.6
tok/s
25–67 · low confidence |
Radeon Instinct MI350X AMD | 288 GB | 8,190 GB/s | Jan 2025 | 70.3 GB | Q8_0 | Comfortable |
|
41.6
tok/s
25–67 · low confidence |
Radeon Instinct MI355X AMD | 288 GB | 8,190 GB/s | Jan 2025 | 70.3 GB | Q8_0 | Comfortable |
|
33.3
tok/s
20–53 · low confidence |
Radeon Instinct MI300 AMD | 128 GB | 6,550 GB/s | Jan 2023 | 70.3 GB | Q8_0 | Comfortable |
|
31.9
tok/s
19–51 · low confidence |
H200 NVL NVIDIA | 141 GB | 4,890 GB/s | Nov 2024 | 70.3 GB | Q8_0 | Comfortable |
|
31.9
tok/s
19–51 · low confidence |
H200 SXM 141 GB NVIDIA | 141 GB | 4,890 GB/s | Nov 2024 | 70.3 GB | Q8_0 | Comfortable |
|
30.5
tok/s
18–49 · low confidence |
Radeon Instinct MI325X AMD | 256 GB | 6,000 GB/s | Oct 2024 | 70.3 GB | Q8_0 | Comfortable |
|
28.1
tok/s
17–45 · low confidence |
GRID A100B NVIDIA | 48 GB | 1,870 GB/s | May 2020 | 40.0 GB | Q4_K_M | Tight |
|
27.4
tok/s
16–44 · low confidence |
A100 PCIe 40 GB NVIDIA | 40 GB | 1,560 GB/s | Jun 2020 | 32.5 GB | Q3_K_M | Tight |
|
27.4
tok/s
16–44 · low confidence |
A100 SXM4 40 GB NVIDIA | 40 GB | 1,560 GB/s | May 2020 | 32.5 GB | Q3_K_M | Tight |
|
27.4
tok/s
16–44 · low confidence |
A800 PCIe 40 GB NVIDIA | 40 GB | 1,560 GB/s | Nov 2022 | 32.5 GB | Q3_K_M | Tight |
|
27.1
tok/s
16–43 · low confidence |
Radeon Instinct MI300A AMD | 128 GB | 5,325 GB/s | Dec 2023 | 70.3 GB | Q8_0 | Comfortable |
|
27.1
tok/s
16–43 · low confidence |
Radeon Instinct MI300X AMD | 192 GB | 5,325 GB/s | Dec 2023 | 70.3 GB | Q8_0 | Comfortable |
|
27.1
tok/s
16–43 · low confidence |
Radeon Instinct MI308X AMD | 192 GB | 5,325 GB/s | Dec 2023 | 70.3 GB | Q8_0 | Comfortable |
|
25.7
tok/s
15–41 · low confidence |
H100 NVL 94 GB NVIDIA | 94 GB | 3,940 GB/s | Mar 2023 | 70.3 GB | Q8_0 | Comfortable |
|
21.9
tok/s
13–35 · low confidence |
H100 PCIe 96 GB NVIDIA | 96 GB | 3,360 GB/s | Mar 2023 | 70.3 GB | Q8_0 | Comfortable |
|
21.9
tok/s
13–35 · low confidence |
H100 SXM5 80 GB NVIDIA | 80 GB | 3,360 GB/s | Oct 2022 | 70.3 GB | Q8_0 | Tight |
|
21.9
tok/s
13–35 · low confidence |
H100 SXM5 94 GB NVIDIA | 94 GB | 3,360 GB/s | Mar 2023 | 70.3 GB | Q8_0 | Comfortable |
|
21.9
tok/s
13–35 · low confidence |
H100 SXM5 96 GB NVIDIA | 96 GB | 3,360 GB/s | Mar 2023 | 70.3 GB | Q8_0 | Comfortable |
|
21.9
tok/s
13–35 · low confidence |
H800 SXM5 NVIDIA | 80 GB | 3,360 GB/s | Mar 2023 | 70.3 GB | Q8_0 | Tight |
|
20.2
tok/s
12–32 · low confidence |
RTX PRO 5000 Blackwell NVIDIA | 48 GB | 1,340 GB/s | Mar 2025 | 40.0 GB | Q4_K_M | Tight |
|
19.1
tok/s
11–31 · low confidence |
H100 SXM5 64 GB NVIDIA | 64 GB | 2,020 GB/s | Mar 2023 | 55.2 GB | Q6_K | Tight |
|
16.7
tok/s
10–27 · low confidence |
Radeon Instinct MI250 AMD | 128 GB | 3,280 GB/s | Nov 2021 | 70.3 GB | Q8_0 | Comfortable |
|
16.7
tok/s
10–27 · low confidence |
Radeon Instinct MI250X AMD | 128 GB | 3,280 GB/s | Nov 2021 | 70.3 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
- XVERSE Technology,Shenzhen Yuanxiang Technology
- Organisation type
- Industry,Industry
- Country
- China
- Published
- 8 December 2023
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Chat, Language modeling/generation
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
- 65B
- Training data
- 2,600,000,000,000 tokens
Based on the name. Exact count unknown but may be listed on Hugging Face.
Training Data: The model has been thoroughly trained on a diversified and high-quality dataset consisting of 2.6 trillion of tokens, including more than 40 languages such as Chinese, English, Russian, and Spanish. The sampling ratio of different types of data is finely set, which makes the performance of Chinese and English excellent, and also takes into account the effect of other languages. Assume 0.85 words per token on average for the mix of languages.
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
- 1.2 × 10²⁴ FLOP
- How it was established
- Operation counting
C = 6ND = 6 * 3.2T tokens * 65B params = 1.248e24 FLOP
The training run
What it physically took to train: which chips, how many, for how long, and what that drew from the wall.
- Wall-clock time
- 4,096 hours (170.7 days)
- Data centre
- There is no paper to reference, no information about hardware used for training found in media.
November 6 to December 8 is 32 days. They did 600B tokens of continual pretraining during this period. The model's total tokens are 3200B. Therefore the total pretraining time was around (32 days * 24 hours/day)*(3200/600) = 4096 hours.
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 source
https://github.com/xverse-ai/XVERSE-65B/blob/main/README_EN.md license info: "The use of the source code in this repository must follow the Apache-2.0 open-source license, while the use of the model weights of XVERSE-65B needs to adhere to the Model License Agreement. The XVERSE-65B model weights are fully open to academic research and support free commercial use. To apply for a commercial license, please fill in the application form. For other questions or collaborations, please contact opens…
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.
- Last updated
- 28 November 2025
The extremes
The ten fastest GPUs that run XVERSE-65B-2
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.
- 01 B300 288 GB · 8,000 GB/s · Q8_0 52.1 tok/s
- 02 B200 180 GB · 8,000 GB/s · Q8_0 52.1 tok/s
- 03 Radeon Instinct MI350X 288 GB · 8,190 GB/s · Q8_0 41.6 tok/s
- 04 Radeon Instinct MI355X 288 GB · 8,190 GB/s · Q8_0 41.6 tok/s
- 05 Radeon Instinct MI300 128 GB · 6,550 GB/s · Q8_0 33.3 tok/s
- 06 H200 NVL 141 GB · 4,890 GB/s · Q8_0 31.9 tok/s
- 07 H200 SXM 141 GB 141 GB · 4,890 GB/s · Q8_0 31.9 tok/s
- 08 Radeon Instinct MI325X 256 GB · 6,000 GB/s · Q8_0 30.5 tok/s
- 09 GRID A100B 48 GB · 1,870 GB/s · Q4_K_M 28.1 tok/s
- 10 A800 PCIe 40 GB 40 GB · 1,560 GB/s · Q3_K_M 27.4 tok/s
The smallest GPUs that still run XVERSE-65B-2
The cheapest route in, by memory capacity. A tight fit runs the model but leaves nothing spare for a longer conversation.
- 01 A800 PCIe 40 GB 40 GB · needs 32.5 GB · Q3_K_M · tight 27.4 tok/s
- 02 A100 PCIe 40 GB 40 GB · needs 32.5 GB · Q3_K_M · tight 27.4 tok/s
- 03 A100 SXM4 40 GB 40 GB · needs 32.5 GB · Q3_K_M · tight 27.4 tok/s
- 04 Radeon PRO W7900D 48 GB · needs 40.0 GB · Q4_K_M · tight 10.1 tok/s
- 05 RTX PRO 5000 Blackwell 48 GB · needs 40.0 GB · Q4_K_M · tight 20.2 tok/s
- 06 RTX 5880 Ada Generation 48 GB · needs 40.0 GB · Q4_K_M · tight 13.0 tok/s
- 07 L20 48 GB · needs 40.0 GB · Q4_K_M · tight 13.0 tok/s
- 08 Radeon PRO W7800 48 GB 48 GB · needs 40.0 GB · Q4_K_M · tight 10.1 tok/s
- 09 Radeon PRO W7900 48 GB · needs 40.0 GB · Q4_K_M · tight 10.1 tok/s
- 10 Data Center GPU Max 1100 48 GB · needs 40.0 GB · Q4_K_M · tight 12.0 tok/s
What the numbers mean
What you need to run it
Minimum card
A100 PCIe 40 GB
Memory needed
32.5 GB
Fastest
52.1 tok/s
XVERSE-65B-2 sits at 65B parameters, which puts it above consumer hardware and into the range where a card is bought for this purpose rather than repurposed for it. 61 of the cards we track can hold it.
The least hardware that works is a A100 PCIe 40 GB. Its 40 GB is enough at Q3_K_M compression, giving roughly 27.4 tokens per second.
At the other end, a B200 generates roughly 52.1 tokens per second on it, on the strength of 8,000 GB/s of memory bandwidth.
Where it came from
XVERSE-65B-2 was published by XVERSE Technology,Shenzhen Yuanxiang Technology, in China, in December 2023. The organisation is categorised as industry,Industry.
It works in Language, and is recorded as doing chat, Language modeling/generation.
Because its weights were released, nothing about running it depends on a provider staying available — it is yours once downloaded.
Understanding the speeds
Across every card that can run it, the middle of the range is about 13.3 tokens per second, and 56 of them clear the ten tokens per second that roughly matches reading speed.
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.
Without the attention layout on record, the memory column is an approximation. It is close enough to choose hardware by, and least reliable at long context.
How it was trained
The training run consumed about 1.2 × 10²⁴ FLOP. That figure describes the cost of creating it and has no bearing on how quickly it generates text.
The training set ran to roughly 2,600,000,000,000 tokens.
Step by step
How to choose a GPU for XVERSE-65B-2
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
Look at what XVERSE-65B-2 actually needs — around 32.5 GB at Q3_K_M. No amount of processing power compensates for a card that cannot hold it.
-
02
Match the context to your actual use
The conversation occupies memory too, and grows as it goes. Set the slider to the length you expect: at long context XVERSE-65B-2 can slip off a card that handles short questions easily.
-
03
Decide how much compression you will accept
Compression is what makes XVERSE-65B-2 fit smaller cards, at some cost in accuracy — Q3_K_M on the smallest card that fits. A minimum quality removes the ones that go too far.
-
04
Sort by speed
Ranking by tokens per second for XVERSE-65B-2 follows memory bandwidth, not core counts, which is why the B200 tops it at 52.1 tok/s.
-
05
Read the fit column last
A tight fit runs XVERSE-65B-2 but leaves nothing spare for a longer conversation; comfortable has headroom. If you expect to grow the context, buy for comfortable.
-
06
See what else that card runs
Following a card through to its own page shows every other model it can hold, which is the question that follows once XVERSE-65B-2 is settled.
Answers
XVERSE-65B-2 — common questions
How accurate are these XVERSE-65B-2 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 31–83 tok/s on the B200 rather than a single number.
What GPU do I need to run XVERSE-65B-2?
The smallest card in our catalogue that holds XVERSE-65B-2 is the A100 PCIe 40 GB, with 40 GB of memory. It runs the model at Q3_K_M using about 32.5 GB, and produces roughly 27.4 tokens per second. 61 cards in total can run it.
How fast is XVERSE-65B-2 on a GPU?
It depends on the card. The quickest we calculate is a B200 at about 52.1 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 56 of the cards that can run XVERSE-65B-2 clear that.
How much VRAM does XVERSE-65B-2 need?
About 32.5 GB at Q3_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.
Is XVERSE-65B-2 open source?
Its weights are published, so XVERSE-65B-2 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.
How many parameters does XVERSE-65B-2 have?
XVERSE-65B-2 has 65B parameters. Based on the name. Exact count unknown but may be listed on Hugging Face. 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.
Who created XVERSE-65B-2?
XVERSE-65B-2 was published by XVERSE Technology,Shenzhen Yuanxiang Technology, based in China, categorised as industry,Industry.
When was XVERSE-65B-2 released?
XVERSE-65B-2 was published in December 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.
What is XVERSE-65B-2 used for?
XVERSE-65B-2 works in Language, and is recorded as handling chat, Language modeling/generation. These are the areas it was designed around; they describe intent rather than a hard boundary.
Where can I download XVERSE-65B-2?
The weights for XVERSE-65B-2 are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.
How much compute was used to train XVERSE-65B-2?
Around 1.2 × 10²⁴ 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.
Can I run XVERSE-65B-2 if it does not fit in my GPU?
It can be split between the card and system memory, but XVERSE-65B-2 generates painfully slowly that way — the nearest miss we calculate is short by 11.2 GB. Nothing on this page assumes offloading.
Would two GPUs run XVERSE-65B-2 faster?
A second card roughly doubles the memory available but not the generation rate. With 61 cards already able to run XVERSE-65B-2 alone, the case for pairing is weak.
Why does the quantisation differ between cards for XVERSE-65B-2?
A larger card holds a more accurate copy. Across the cards that run XVERSE-65B-2, 4 compression levels are used; the floor control above pins it to one.
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.