Skip to main content

DeepSeek V3.2

An open-weight 685B mixture-of-experts model with strong reasoning at a low price.

DeepSeek V3.2 specifications

ProviderDeepSeek
Context window163,840 tokens (164K)
Input price$0.28 / 1M tokens
Output price$0.42 / 1M tokens
InputsText
Reasoning modeYes
Vision (image input)No
Tool / function callingYes
Native real-time webNo
Open-weightYes
ReleasedDecember 2025

Pricing & context web-verified June 2026 confirm on DeepSeek's pricing page. Context-cache hits bill far less; V4 tiers also available.

What DeepSeek V3.2 is best for

  • Cost-sensitive reasoning and coding
  • Self-hosted / open-weight deployments
  • Math and code

How to prompt DeepSeek V3.2

  • Text-only — use a multimodal model if you need image input.
  • It reasons while calling tools ("thinking in tool-use") — give it tools for agentic tasks.
  • Open weights mean you can self-host; cache context to cut cost further.

Work with DeepSeek V3.2

Frequently asked questions

What is DeepSeek V3.2's context window?

DeepSeek V3.2 has a 163,840-token context window (about 164K tokens) — the maximum amount of text it can consider at once.

How much does DeepSeek V3.2 cost?

Via the DeepSeek API, DeepSeek V3.2 costs $0.28 per 1M input tokens and $0.42 per 1M output tokens (web-verified June 2026). Output tokens cost more, so the length of the model's reply usually drives the bill.

Is DeepSeek V3.2 multimodal?

No — DeepSeek V3.2 is text-only. Use a vision-capable model if you need to send images.

Is DeepSeek V3.2 open-weight?

Yes — DeepSeek V3.2 is open-weight, so you can self-host it or run it through a third-party host.