DeepSeek V3.2 API
About
Build advanced reasoning agents with DeepSeek V3.2, a specialized AI model optimized for complex logic and chain-of-thought tasks. This model delivers available completely free of charge, native tool calling support. Access DeepSeek V3.2 via the Nvidia API with up to 66K output tokens.
Capabilities
Input Modalities
text
Output Modalities
text
Reasoning
Yes
Structured Output
No
Tool Use
Yes
WeightsClosed
Temperature
Adjustable
Attachment
Not Supported
Limits
Context Window
164K
Tokens
Input Limit
164K
Tokens
Max Output
66K
Tokens
Pricing
Standard (per 1M tokens)
Input
Free
Output
Free
Frequently Asked Questions
DeepSeek V3.2 by Nvidia costs Free per 1M input tokens and Free per 1M output tokens.
Model Details
ID
deepseek-ai/deepseek-v3.2
Provider
NvidiaFamily
deepseek
Release Date
Dec 1, 2025
Knowledge Cutoff
Jul 1, 2024
API Integration
NPM Package
@ai-sdk/openai-compatible
Environment Variables
NVIDIA_API_KEY
API Base URL
https://integrate.api.nvidia.com/v1