DeepSeek: DeepSeek V3.2

deepseek/deepseek-v3.2

DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...

Context window: 163,840 tokens
Input: text
Output: text
Relay support: Text & documents

Open in workspace →OpenRouter model page

Reference prices & points

Our global markup: 15% · 1 USD = 10,000 points · Price version 0

Billing unit	OpenRouter / USD	Our reference / USD	Points
Input / 1M tokens	$0.26	$0.299	2,990
Output / 1M tokens	$0.38	$0.437	4,370
Cache read / 1M tokens	$0.13	$0.1495	1,495

Actual charge = the OpenRouter request cost × the markup locked for that request. Catalogue rates may be routing references; caching, long context, images and audio are charged when used. Listed prices do not mean every capability is available through this relay.

Catalogue time: 2026-11-11T23:10:25.766Z · Live catalogue

View all upstream price fields

{
  "prompt": "0.00000026",
  "completion": "0.00000038",
  "input_cache_read": "0.00000013"
}

Integration & availability

Use /v1/chat/completions for non-streaming, multi-turn conversations. Models marked for vision accept images; documents can be extracted to text in the workspace. Create a site API key and ensure you have enough points and the service is enabled. Provider availability, context and budget limits apply.

Read the API guide → · Back to model directory