DeepSeek V3.2
Released 2025-12 · hybrid · 164K tokens · 6 benchmarks · Open weight
Editorial notes
671B totales, 37B activos. Sparse Attention. Scores reflejan modo Reasoning (thinking-on) per paper oficial. AA-Index 42 corresponde a variante Reasoning (32 para Non-reasoning).
Spec sheet
- Company
- DeepSeek
- Country
- CN
- Type
- hybrid
- Release
- 2025-12
- Context
- 164K tokens
- Params total
- 671B
- Params active (MoE)
- 37B
- License
- deepseek-license
- Quants
- BF16, Q8_0, Q5_K_M, Q4_K_M, Q3_K_M
- Pricing (deepseek)
- $0.27/$1.1/M
- Slug
- deepseek-v3-2
Quick install
3 toolsollama run deepseek-v3.2Note: Q4_K_M ~390 GB VRAM
Buscar "DeepSeek-V3.2" en LM Studiovllm serve deepseek-ai/DeepSeek-V3.2 --tensor-parallel-size 8Pricing comparison
| Provider | Input/MTok | Output/MTok | As of |
|---|---|---|---|
| deepseek | $0.27 | $1.10 | 2025-12-01 |
| openrouter | $0.40 | $1.50 | 2025-12-01 |
Benchmarks (6)
Reasoning 2
Coding 3
Cite this model
BibTeX · APA
BibTeX
@misc{frontier-deepseek-v3-2,
title = {DeepSeek V3.2},
author = {{DeepSeek}},
year = {2025},
note = {Frontier Benchmarks AI atlas. Accessed 2026-05-08},
url = {https://frontierbenchmarks.com/models/deepseek-v3-2}
} APA
DeepSeek (2025). DeepSeek V3.2 [Large language model]. Frontier Benchmarks AI. Retrieved 2026-05-08, from https://frontierbenchmarks.com/models/deepseek-v3-2
Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.