DeepSeek API pricing

DeepSeek V3.2 API pricing

DeepSeek V3.2 costs $0.28 per million input tokens and $0.40 per million output tokens on the DeepSeek API, and $0.028 per million cached input tokens. Context window: 164k tokens. Updated .

Price per 1M tokensUS dollars
Input$0.28
Cached input$0.028
Output$0.40

What it costs in practice

  • One chat or extraction call (2,000 tokens in, 500 out): $0.0008
  • 1,000 such calls: $0.760
  • 1 million tokens at a typical 80% input / 20% output mix: $0.304

Abilities: tool use, reasoning. Batch, long-context, priority and regional prices differ; check the full table and the provider’s own pricing page before you commit.

Cheapest alternatives from other providers

ModelInputOutputContextTypical call ×1,000
Qwen Turbo
Alibaba (Qwen)
$0.050$0.20129k$0.20
GPT-5 nano
OpenAI
$0.050$0.40272k$0.30
Gemini 2.5 Flash Lite
Google
$0.10$0.401.0M$0.40
Claude Haiku 5.5
Anthropic
$0.10$0.501M$0.45
GPT-6 luna
OpenAI
$0.10$0.50922k$0.45
Qwen3.8 Flash
Alibaba (Qwen)
$0.15$0.47992k$0.54

Other DeepSeek models

ModelInputOutputContextTypical call ×1,000
DeepSeek Coder
DeepSeek
$0.14$0.28128k$0.42
DeepSeek V3
DeepSeek
$0.27$1.1066k$1.09
DeepSeek Flash
DeepSeek
$0.30$1.201M$1.20
DeepSeek V4 Flash
DeepSeek
$0.30$1.201M$1.20
DeepSeek V4 Flash Vision Exp
DeepSeek
$0.30$1.201M$1.20
DeepSeek R1
DeepSeek
$0.55$2.1966k$2.19

Questions

How much does DeepSeek V3.2 cost?

DeepSeek V3.2 costs $0.28 per million input tokens and $0.40 per million output tokens on the DeepSeek API; cached input is $0.028 per million. A typical call with 2,000 tokens in and 500 out costs about $0.0008.

What is the context window of DeepSeek V3.2?

DeepSeek V3.2 accepts up to 163,840 input tokens and returns up to 163,840 output tokens.

Is there a cheaper alternative to DeepSeek V3.2?

For a typical call, Qwen Turbo (Alibaba (Qwen), $0.0002 per call), GPT-5 nano (OpenAI, $0.0003 per call), Gemini 2.5 Flash Lite (Google, $0.0004 per call) cost less and also support tool use, reasoning. Quality differs, so test on your own prompts.

Source: the open LiteLLM model price map (MIT), checked daily (data file).