DeepSeek-V2 (MoE-236B) TPS calculator

Open weights DeepSeek 236B parameters May 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

17 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Radeon Instinct MI250

128 GB · Q3_K_M · 68.8 tok/s

Fastest card

B200

142 tok/s · 180 GB

Which GPUs can run DeepSeek-V2 (MoE-236B)?

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

17 cards match

Calculating

					Needs	Quantisation	Fit
142 tok/s 85–228 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	157.4 GB	Q5_K_M	Tight
137 tok/s 82–220 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	102.5 GB	Q3_K_M	Tight
120 tok/s 72–192 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	116.2 GB	IQ4_XS	Tight
120 tok/s 72–192 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	116.2 GB	IQ4_XS	Tight
112 tok/s 67–179 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	102.5 GB	Q3_K_M	Tight
79.8 tok/s 48–128 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	239.9 GB	Q8_0	Tight
74.0 tok/s 44–118 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	157.4 GB	Q5_K_M	Tight
74.0 tok/s 44–118 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	157.4 GB	Q5_K_M	Tight
68.8 tok/s 41–110 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	102.5 GB	Q3_K_M	Tight
68.8 tok/s 41–110 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	102.5 GB	Q3_K_M	Tight
67.8 tok/s 41–108 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	184.9 GB	Q6_K	Comfortable
63.7 tok/s 38–102 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	239.9 GB	Q8_0	Tight
63.7 tok/s 38–102 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	239.9 GB	Q8_0	Tight
57.4 tok/s 34–92 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	102.5 GB	Q3_K_M	Tight
56.1 tok/s 34–90 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	102.5 GB	Q3_K_M	Tight
7.4 tok/s 4–12 · low confidence	GB10 NVIDIA	128 GB	273 GB/s	Oct 2025	102.5 GB	Q3_K_M	Tight
7.4 tok/s 4–12 · low confidence	Jetson T5000 NVIDIA	128 GB	273 GB/s	Aug 2025	102.5 GB	Q3_K_M	Tight

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: DeepSeek
Organisation type: Industry
Country: China
Published: 7 May 2024
Authors: DeepSeek-AI, Aixin Liu, Bei Feng, Bin Wang, Bingxuan Wang, Bo Liu, Chenggang Zhao, Chengqi Dengr, Chong Ruan, Damai Dai, Daya Guo, Dejian Yang, Deli Chen, Dongjie Ji, Erhang Li, Fangyun Lin, Fuli Luo, Guangbo Hao, Guanting Chen, Guowei Li, H. Zhang, Hanwei Xu, Hao Yang, Haowei Zhang, Honghui Ding, Huajian Xin, Huazuo Gao, Hui Li, Hui Qu, J.L. Cai, Jian Liang, Jianzhong Guo, Jiaqi Ni, Jiashi Li, Ji…

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

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: 236B
Training data: 8,100,000,000,000 tokens
Batch size: 18,432,000

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: 1 × 10²⁴ 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 H800 SXM5
Chip-hours: 172,800
Data centre: Paper on DeepSeek-V3

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: deepseek-ai

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: DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
Last updated: 11 February 2026

The extremes

The ten fastest GPUs that run DeepSeek-V2 (MoE-236B)

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 DeepSeek-V2 (MoE-236B)

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

Radeon Instinct MI250

Memory needed

102.5 GB

Fastest

142 tok/s

DeepSeek-V2 (MoE-236B) reaches a parameter count of 236B. That is beyond what any single graphics card holds. Running it means either splitting it across several cards or renting hardware built for the job, and every card able to hold it alone is a datacentre part. The number that can: 17.

The entry point is Radeon Instinct MI250, with a memory capacity of 128 GB, running it at a compression of Q3_K_M and producing around 68.8 tokens per second.

The quickest result comes from B200, generating roughly 142 tokens per second on the strength of a memory bandwidth of 8,000 GB/s.

What this model is

DeepSeek-V2 (MoE-236B) was published by DeepSeek, in the country recorded as China, during May 2024. The publishing organisation is categorised as industry.

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

The weights being open is what puts this page in the calculator rather than only in the catalogue: it is a model you can actually hold. On Hugging Face it is published under the organisation deepseek-ai.

What decides the speed

The median result is around 68.8 tokens per second. Clearing the ten tokens per second that roughly matches reading speed: 15 of them.

This is a mixture-of-experts model, which routes each token through only part of itself. It therefore generates far faster than its total size suggests — while still needing every parameter resident in memory, so it is quick without being cheap to hold.

Its internal architecture is not on file, so memory is approximated from the parameter count and marked accordingly. Expect the real figure to differ, more so at long context.

How it was trained

Training it took a computation budget of roughly 1 × 10²⁴ FLOP, on hardware recorded as NVIDIA H800 SXM5. That figure measures what producing the model cost, and has no bearing on how fast it answers.

The training set ran to roughly 8,100,000,000,000 tokens of text.

Step by step

How to choose a GPU for DeepSeek-V2 (MoE-236B)

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

Start from what it actually needs, which is the requirement of DeepSeek-V2 (MoE-236B), needing around 102.5 GB at a compression of Q3_K_M. No amount of processing power compensates for a card that cannot hold it.
02

Set the context length you will work at

Longer conversations cost memory on top of what the weights need. Move the slider to your real working length before trusting any row for DeepSeek-V2 (MoE-236B).
03

Set a quality floor

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

Sort by speed to see how cards rank for DeepSeek-V2 (MoE-236B). It will not match a gaming ordering, because generation is bound by memory bandwidth. The card topping the list is B200, at 142 tok/s.
05

Check the fit verdict before buying

Tight means it loads and works with no room to raise the context later, in the case of DeepSeek-V2 (MoE-236B). 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 DeepSeek-V2 (MoE-236B).

Answers

DeepSeek-V2 (MoE-236B) — common questions

DeepSeek-V2 (MoE-236B)— 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.

DeepSeek-V2 (MoE-236B)— how many parameters does it have?

It has a parameter count of 236B. 21B active params, 236B total. 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.

DeepSeek-V2 (MoE-236B)— who created it?

It was published by DeepSeek, based in China, an organisation categorised as industry.

DeepSeek-V2 (MoE-236B)— when was it released?

It was published in May 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.

DeepSeek-V2 (MoE-236B)— 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. These are the areas it was designed around; they describe intent rather than a hard boundary.

DeepSeek-V2 (MoE-236B)— where can I download it?

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

DeepSeek-V2 (MoE-236B)— how much compute was used to train it?

Training consumed around 1 × 10²⁴ FLOP, on hardware recorded as NVIDIA H800 SXM5. 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.

DeepSeek-V2 (MoE-236B)— 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 43.6 GB. Every figure here assumes the whole model is resident on the card.

DeepSeek-V2 (MoE-236B)— would two GPUs run it faster?

A second card roughly doubles the memory available but not the generation rate. The number already able to run it alone: 17. So a second card is rarely the answer here.

DeepSeek-V2 (MoE-236B)— why does the quantisation differ between cards?

A larger card holds a more accurate copy. The number of compression levels used across the cards that run it: 5. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.

DeepSeek-V2 (MoE-236B)— how accurate are these speed estimates?

These are estimates with real error bars, and any of them could reasonably land anywhere in its published range depending on which runtime you use. The fastest result here: 85–228 tok/s on B200. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

DeepSeek-V2 (MoE-236B)— what GPU do I need to run it?

The smallest card in our catalogue that holds it is Radeon Instinct MI250, with a memory capacity of 128 GB. It runs the model at a compression of Q3_K_M using about 102.5 GB, and produces roughly 68.8 tokens per second. The number of cards able to run it in total: 17.

DeepSeek-V2 (MoE-236B)— how fast is it on a GPU?

It depends on the card. The quickest we calculate is B200, at about 142 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: 15.

DeepSeek-V2 (MoE-236B)— how much VRAM does it need?

It needs about 102.5 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.

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