DeepSeek-V3.2-Exp TPS calculator

Open weights DeepSeek 671B parameters September 2025

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

0 cards that can run it

818 cards we hold specifications for

Which GPUs can run DeepSeek-V3.2-Exp?

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.

0 cards match

Calculating
Needs Quantisation Fit

No card in our catalogue can run this model with these settings.

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
29 September 2025

What it does

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

Domain
Language
Task
Language modeling/generation, Code generation, Quantitative reasoning, Question answering, Search, System control, Instruction interpretation
Base model
DeepSeek-V3.1-Terminus

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
671B

671B total, 37B active (assuming same as in the base model)

Training data
943,700,000,000 tokens

"We train both the main model and the indexer for 15000 steps, with each step consisting of 480 sequences of 128K tokens, resulting in a total of 943.7B 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
4.2 × 10²⁴ FLOP

3.594058e+24 FLOP [base model] + 2.095014e+23 FLOP = 3.8035594e+24 FLOP for pre-training plus continued pre-training Then, "this framework allocates a post-training computational budget exceeding 10% of the pre-training cost", so 4.18e24

How it was established
Operation counting
Fine-tuning compute
5.9 × 10²³ FLOP

Continued pretraining compute: 6 * 37B * 943.7B = 2.095e23 FLOP. In addition, V3.2 states post-training compute budget exceeds 10% of total pretraining cost, implying additional post-training FLOPs > 3.8e23 (so finetune total beyond V3.1-Terminus is >5.9e23 FLOP).

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)
Training code
Unreleased

MIT license https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp

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.

Record confidence
Confident

Sources

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

Reference
Introducing DeepSeek-V3.2-Exp
Last updated
19 February 2026

What the numbers mean

What you need to run it

At 671B parameters, DeepSeek-V3.2-Exp is beyond what any single graphics card holds. Running it means either splitting it across several cards or renting hardware built for the job — 0 of the cards we track can hold it on their own, and all of them are datacentre parts.

What this model is

DeepSeek-V3.2-Exp was published by DeepSeek, in China, in September 2025. The organisation is categorised as industry.

It works in Language, and is recorded as doing language modeling/generation, Code generation, Quantitative reasoning, Question answering, Search, System control, Instruction interpretation.

Its starting point was DeepSeek-V3.1-Terminus — most models at this scale are adapted from an existing base rather than built from nothing.

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. It is published under the deepseek-ai organisation on Hugging Face.

How it was trained

The training run consumed about 4.2 × 10²⁴ FLOP. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

It was trained on about 943,700,000,000 tokens of text.

Step by step

How to choose a GPU for DeepSeek-V3.2-Exp

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

    The table lists every card that can hold DeepSeek-V3.2-Exp. That figure, not the card's headline performance, is what decides whether it runs.

  2. 02

    Decide how long your conversations run

    Set the context to what you will actually use. The cache grows with the conversation, and it is the usual reason DeepSeek-V3.2-Exp stops fitting a card that seemed fine.

  3. 03

    Set a quality floor

    Each card runs the least-compressed copy it can hold. Setting a floor drops the cards that only manage DeepSeek-V3.2-Exp by squeezing it further than you would want.

  4. 04

    Rank by throughput rather than spec sheet

    Ranking by tokens per second for DeepSeek-V3.2-Exp follows memory bandwidth, not core counts.

  5. 05

    Check the fit verdict before buying

    Tight means DeepSeek-V3.2-Exp loads and works, with no room to raise the context later. Comfortable means you can. The difference matters more than a few tokens per second.

  6. 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. Worth a look before buying for DeepSeek-V3.2-Exp alone — a card is usually bought for more than one model.

Answers

DeepSeek-V3.2-Exp — common questions

01

How many parameters does DeepSeek-V3.2-Exp have?

DeepSeek-V3.2-Exp has 671B parameters. 671B total, 37B active (assuming same as in the base model). 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.

02

Who created DeepSeek-V3.2-Exp?

DeepSeek-V3.2-Exp was published by DeepSeek, based in China, categorised as industry.

03

When was DeepSeek-V3.2-Exp released?

DeepSeek-V3.2-Exp was published in September 2025.

04

What is DeepSeek-V3.2-Exp used for?

DeepSeek-V3.2-Exp works in Language, and is recorded as handling language modeling/generation, Code generation, Quantitative reasoning, Question answering, Search, System control, Instruction interpretation. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.

05

Where can I download DeepSeek-V3.2-Exp?

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

06

How much compute was used to train DeepSeek-V3.2-Exp?

Around 4.2 × 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.

07

Can I run DeepSeek-V3.2-Exp 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 DeepSeek-V3.2-Exp is rarely worth using — the nearest miss we calculate is short by 103.0 GB. Every figure here assumes the whole model is on the card.

08

Would two GPUs run DeepSeek-V3.2-Exp faster?

A second card roughly doubles the memory available but not the generation rate. With 0 cards already able to run DeepSeek-V3.2-Exp alone, the case for pairing is weak.

09

Why does the quantisation differ between cards for DeepSeek-V3.2-Exp?

Because capacity varies, so does how hard DeepSeek-V3.2-Exp has to be squeezed — 1 distinct levels appear in the table above. Set a minimum quality to compare at one.

10

How accurate are these DeepSeek-V3.2-Exp speed estimates?

These are estimates with real error bars. The fastest result here, the range beneath each figure, could reasonably land anywhere in its published range depending on which runtime you use.

11

Is DeepSeek-V3.2-Exp open source?

Its weights are published, so DeepSeek-V3.2-Exp 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 19 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.