DeepSeek V3.2

DeepSeek

DeepSeek V3.2 with 685B total parameters (37B active). Features Multi-Token Prediction and Sparse Attention for efficiency.

Text Generation Local Latest DeepSeek Family vV3.2
Documentation Back to Models
Parameters
685B
params
Context Window
128K
tokens
Max Output
8K
tokens
Input Price
-
per 1M tokens
Output Price
-
per 1M tokens

DeepSeek Family Timeline 13 versions

DeepSeek V3.2 Current Latest local
Dec 2025
DeepSeek V3.2 Exp Latest api
Sep 2025
DeepSeek V3.1 api
Aug 2025
DeepSeek R1 (May) api
May 2025
DeepSeek V3 (March) api
Mar 2025
DeepSeek R1 Distill 7B local
Jan 2025
DeepSeek R1 Distill 32B local
Jan 2025
DeepSeek R1 api
Jan 2025
DeepSeek Coder 6.7B local
Nov 2024
DeepSeek Coder 33B local
Nov 2024
DeepSeek V2 Lite 16B local
May 2024
DeepSeek Chat api
-
DeepSeek Coder api
-

Capabilities

👁️
Vision
⚡
Function Calling
📋
JSON Mode
🌊
Streaming
💬
System Prompt
🖥️
Code Execution
🔍
Web Search
🔌
MCP Support

Local Model Specs

Architecture
MoE with Sparse Attention

Details

Release Date
December 1, 2025
Knowledge Cutoff
-
Source
Local
License
MIT
Model ID
deepseek-v3-2
Last updated: December 28, 2025