Title: DeepSeek-V3.2 %E2%80%94 deepseek-v3-2
Canonical: https://llmeu.com/cs/models/deepseek-v3-2/
Markdown mirror: https://llmeu.com/cs/models/deepseek-v3-2.md
Machine-readable: https://llmeu.com/llms.txt (index) - https://llmeu.com/llms-full.txt (full catalogue) - https://llmeu.com/llms-catalogue.json
Provider: DeepSeek
Licence: mit
Hosting: catalogue only %E2%80%94 not hosted on LLM EU yet
Sources: https://huggingface.co/deepseek-ai/DeepSeek-V3.2

# DeepSeek-V3.2

[DeepSeek](/cs/providers/deepseek/)

Naposledy zkontrolov%C3%A1no: 14. 10. 2026

## Suverenita a hosting

Tento model je uveden v katalogu pro referenci. Zat%C3%ADm nen%C3%AD hostov%C3%A1n na LLM EU a %C5%BE%C3%A1dn%C3%BD endpoint prvn%C3%AD strany zde neposkytuje jeho slu%C5%BEby. Po%C5%BEadavky na n%C4%9Bj sm%C4%9B%C5%99uj%C3%AD p%C5%99%C3%ADmo k poskytovateli, podle jeho vlastn%C3%ADch podm%C3%ADnek.

## Shrnut%C3%AD

DeepSeek-V3.2 je model typu mixture-of-experts s 685 miliardami parametr%C5%AF, vydan%C3%BD pod licenc%C3%AD MIT, s konfigura%C4%8Dn%C3%ADm maximem 163 840 token%C5%AF. Jeho hlavn%C3%AD architektonickou zm%C4%9Bnou je DeepSeek Sparse Attention, kterou dodavatel popisuje jako sn%C3%AD%C5%BEen%C3%AD v%C3%BDpo%C4%8Detn%C3%ADch n%C3%A1klad%C5%AF na inferenci s dlouh%C3%BDm kontextem. Jedn%C3%A1 se o model pouze pro text a vy%C5%BEaduje nasazen%C3%AD na v%C3%ADce uzlech s plnou p%C5%99esnost%C3%AD.

## Specifikace

- Poskytovatel
  DeepSeek (CN)
- Kontextov%C3%A9 okno
  164k
- Parametry
  685.36B
- Modality
  text
- Licence
  mit %C2%B7 [licence](https://opensource.org/license/mit)
- Otev%C5%99en%C3%A9 v%C3%A1hy
  yes
- Datum vyd%C3%A1n%C3%AD
  01. 12. 2025
- %C3%9Alohy
  [Uva%C5%BEov%C3%A1n%C3%AD](/cs/tasks/reasoning/), [K%C3%B3dov%C3%A1n%C3%AD](/cs/tasks/coding/), [Shrnut%C3%AD](/cs/tasks/summarization/), [Vyhled%C3%A1v%C3%A1n%C3%AD (RAG)](/cs/tasks/rag/)

## Jazyky

**Siln%C3%A1:** en, zh

**Dostate%C4%8Dn%C3%A1:** de, fr, es, it, pt, ru, ja, ko, ar

#### Siln%C3%A1

- en
- zh

#### Dostate%C4%8Dn%C3%A1

- de
- fr
- es
- it
- pt
- ru
- ja
- ko
- ar

Neov%C4%9B%C5%99eno: poskytovatel toto tvrzen%C3%AD nedokumentoval. %E2%80%94 a language is listed as strong only where a source claims it; the card links the evidence where one exists.

## Kdy pou%C5%BE%C3%ADt

#### Siln%C3%A9 str%C3%A1nky

- MIT licence
- sparse attention aimed at lower long-context inference cost
- 163,840-token configuration maximum

#### Omezen%C3%AD

- 685B parameters: full-precision serving needs several nodes
- vendor efficiency claims are not independently reproduced here
- no evaluation has been run by LLM EU

## %C4%8Casto kladen%C3%A9 ot%C3%A1zky

### Is DeepSeek-V3.2 served on LLM EU?

It has no endpoint in the catalogue, so it is catalogue only %E2%80%94 not hosted on LLM EU yet, with no region and no retention to report. A card whose status is "listed" is a catalogue entry, not a running deployment.

### Under what licence is DeepSeek-V3.2 released?

DeepSeek-V3.2 is recorded under mit. The recorded licence text is https://opensource.org/license/mit. The weights are downloadable (open_weights is true). LLM EU republishes the licence id and the link and gives no legal advice; the licence text governs.

### How large is DeepSeek-V3.2, and what context window does it have?

DeepSeek-V3.2 is recorded at 685B (safetensors total 685,355,329,792) (source: https://huggingface.co/api/models/deepseek-ai/DeepSeek-V3.2), with a context window of 163840 tokens (source: https://huggingface.co/deepseek-ai/DeepSeek-V3.2/blob/main/config.json).

### Ceny

Tento model je uveden v katalogu pro referenci. Zat%C3%ADm nen%C3%AD hostov%C3%A1n na LLM EU a %C5%BE%C3%A1dn%C3%BD endpoint prvn%C3%AD strany zde neposkytuje jeho slu%C5%BEby. Po%C5%BEadavky na n%C4%9Bj sm%C4%9B%C5%99uj%C3%AD p%C5%99%C3%ADmo k poskytovateli, podle jeho vlastn%C3%ADch podm%C3%ADnek.

### Stav ov%C4%9B%C5%99en%C3%AD

Pokud pole nem%C3%A1 zdroj, karta jej zobraz%C3%AD jako neov%C4%9B%C5%99en%C3%A9, m%C3%ADsto aby jej vyplnila. Pr%C3%A1zdn%C3%A9 pole je informace.

LLM EU Verified znamen%C3%A1 zdokumentov%C3%A1no, nikoli pr%C3%A1vn%C4%9B certifikov%C3%A1no. Nejde o posouzen%C3%AD shody podle AI Act a LLM EU nen%C3%AD org%C3%A1nem EU.

### Verze a varianty

- catalogue %C2%B7 Pouze katalog %C2%B7 [weights](https://huggingface.co/deepseek-ai/DeepSeek-V3.2)

### P%C5%99%C3%ADklad po%C5%BEadavku

```
curl https://api.llmeu.com/v1/chat/completions \
  -H "Authorization: Bearer $LLMEU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek-v3-2","messages":[{"role":"user","content":"Hello"}]}'
```

Pro tento model nen%C3%AD nasazen %C5%BE%C3%A1dn%C3%BD playground: %C5%BE%C3%A1dn%C3%BD koncov%C3%BD bod jej neobsluhuje, tak%C5%BEe po%C5%BEadavek by nem%C4%9Bl kam j%C3%ADt.

### Benchmarky

%C5%BD%C3%A1dn%C3%A9 zve%C5%99ejn%C4%9Bn%C3%A9 hodnocen%C3%AD. Toto je mezera, nikoli skryt%C3%BD v%C3%BDsledek.

### Zdroje

- [huggingface.co](https://huggingface.co/deepseek-ai/DeepSeek-V3.2)

### Historie zm%C4%9Bn karty

- 01. 12. 2025 %E2%80%94 Datum vyd%C3%A1n%C3%AD
- 14. 10. 2026 %E2%80%94 Naposledy zkontrolov%C3%A1no

Naposledy zkontrolov%C3%A1no: 14. 10. 2026