Title: DeepSeek-V3.2 %E2%80%94 deepseek-v3-2
Canonical: https://llmeu.com/ro/models/deepseek-v3-2/
Markdown mirror: https://llmeu.com/ro/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](/ro/providers/deepseek/)

Ultima revizuire: 14 oct. 2026

## Suveranitate %C8%99i g%C4%83zduire

Acest model este catalogat ca referin%C8%9B%C4%83. Nu este g%C4%83zduit %C3%AEnc%C4%83 pe LLM EU %C8%99i niciun endpoint propriu nu %C3%AEl serve%C8%99te aici. Cererile c%C4%83tre el merg direct la furnizor, conform termenilor acelui furnizor.

## Rezumat

DeepSeek-V3.2 este un model mixture-of-experts de 685B lansat sub licen%C8%9Ba MIT, cu un maxim de configura%C8%9Bie de 163.840 de token-uri. Principala sa modificare arhitectural%C4%83 este DeepSeek Sparse Attention, pe care furnizorul o descrie ca reduc%C3%A2nd costul computa%C8%9Bional al inferen%C8%9Bei cu context lung. Este un model doar text %C8%99i necesit%C4%83 o implementare multi-nod la precizie complet%C4%83.

## Specifica%C8%9Bii

- Furnizor
  DeepSeek (CN)
- Fereastr%C4%83 de context
  164k
- Parametri
  685.36B
- Modalit%C4%83%C8%9Bi
  text
- Licen%C8%9B%C4%83
  mit %C2%B7 [licence](https://opensource.org/license/mit)
- Ponderi deschise
  yes
- Data lans%C4%83rii
  01 dec. 2025
- Sarcini
  [Ra%C8%9Bionament](/ro/tasks/reasoning/), [Codare](/ro/tasks/coding/), [Sumarizare](/ro/tasks/summarization/), [Reg%C4%83sire (RAG)](/ro/tasks/rag/)

## Limbi

**Puternic:** en, zh

**Adecvat:** de, fr, es, it, pt, ru, ja, ko, ar

#### Puternic

- en
- zh

#### Adecvat

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

Neverificat: furnizorul nu a documentat aceast%C4%83 afirma%C8%9Bie. %E2%80%94 a language is listed as strong only where a source claims it; the card links the evidence where one exists.

## C%C3%A2nd se folose%C8%99te

#### Puncte forte

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

#### Limit%C4%83ri

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

## %C3%8Entreb%C4%83ri frecvente

### 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).

### Pre%C8%9Buri

Acest model este catalogat ca referin%C8%9B%C4%83. Nu este g%C4%83zduit %C3%AEnc%C4%83 pe LLM EU %C8%99i niciun endpoint propriu nu %C3%AEl serve%C8%99te aici. Cererile c%C4%83tre el merg direct la furnizor, conform termenilor acelui furnizor.

### Stare de verificare

Acolo unde un c%C3%A2mp nu are surs%C4%83, cardul %C3%AEl afi%C8%99eaz%C4%83 ca neverificat %C3%AEn loc s%C4%83 %C3%AEl completeze. Un c%C3%A2mp gol este informa%C8%9Bie.

LLM EU Verified %C3%AEnseamn%C4%83 documentat, nu certificat legal. Nu este o evaluare a conformit%C4%83%C8%9Bii cu AI Act %C8%99i LLM EU nu este un organism al UE.

### Versiuni %C8%99i variante

- catalogue %C2%B7 Doar %C3%AEn catalog %C2%B7 [weights](https://huggingface.co/deepseek-ai/DeepSeek-V3.2)

### Exemplu de cerere

```
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"}]}'
```

Nu exist%C4%83 un playground implementat pentru acest model: niciun punct final nu %C3%AEl serve%C8%99te, a%C8%99a c%C4%83 o cerere nu ar avea unde s%C4%83 ajung%C4%83.

### Benchmark-uri

Nicio evaluare publicat%C4%83. Aceasta este o lacun%C4%83, nu un rezultat ascuns.

### Surse

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

### Istoric modific%C4%83ri fi%C8%99%C4%83

- 01 dec. 2025 %E2%80%94 Data lans%C4%83rii
- 14 oct. 2026 %E2%80%94 Ultima revizuire

Ultima revizuire: 14 oct. 2026