DeepSeek-V3.2-Exp

Open weights Active Text LLM MoE deepseek-v3-2-exp

Experimental V3.1-Terminus derivative introducing DeepSeek Sparse Attention (DSA) for long context.

Released
Sep 29, 2025
Context window
164K tokens
Max output
8.2K tokens
Parameters
671B 37B active
Input price
Not disclosed
Output price
Not disclosed

Prices are first-party list prices in USD at the standard tier. Last verified Oct 30, 2026.

Pricing

Launch priceSep 29, 2025
$0.28 in / $0.42 out
Current price
No longer sold

Specifications

Developer
DeepSeek
Model type
Text LLM
Architecture
MoE
Parameters
671B (37B active)
Input
Text
Output
Text
Knowledge cutoff
Not disclosed
License
MIT License
API access
No first-party API
Capabilities
Long context Reasoning