Home/Models/DeepSeek V3.2 (Reasoner)

DeepSeek V3.2 (Reasoner)

Open Source
by DeepSeek· Released 2025-12
Official Docs ↗

DeepSeek V3.2 reasoner is the thinking-mode variant of DeepSeek V3.2, delivering reasoning performance competitive with o3 and Claude Opus at roughly 1/10th the price. It supports up to 64K output tokens and is the strongest cheap option for math and coding reasoning.

Input Price
$0.28/1M
per 1M tokens
Output Price
$0.42/1M
per 1M tokens
Context Window
128K tokens
max tokens
ELO Score
1385
LMSYS Arena

Pricing Breakdown

VolumeInput CostOutput CostCombined (50/50)
1,000 tokens$0.0003$0.0004$0.0003
10,000 tokens$0.0028$0.0042$0.0035
100,000 tokens$0.0280$0.0420$0.0350
1,000,000 tokens$0.2800$0.4200$0.3500

Strengths

  • o-series-class reasoning at a fraction of the price
  • Up to 64K output tokens for long chain-of-thought
  • Open weights
  • Cache-hit pricing available

Weaknesses

  • Verbose reasoning increases effective output cost
  • 128K context vs. 200K on US competitors

Best For

reasoningmathcodingresearch

Compare DeepSeek V3.2 (Reasoner) With

DeepSeek V3.2 (Reasoner) vs GPT-5.4
OpenAI
DeepSeek V3.2 (Reasoner) vs Claude Opus 4.7
Anthropic
DeepSeek V3.2 (Reasoner) vs Gemini 3.1 Pro
Google
DeepSeek V3.2 (Reasoner) vs o3
OpenAI
DeepSeek V3.2 (Reasoner) vs Claude Sonnet 4.6
Anthropic
DeepSeek V3.2 (Reasoner) vs Gemini 2.5 Pro
Google

Frequently Asked Questions

How much does DeepSeek V3.2 (Reasoner) cost?

DeepSeek V3.2 (Reasoner) costs $0.28/1M per 1M input tokens and $0.42/1M per 1M output tokens.

What is DeepSeek V3.2 (Reasoner)'s context window?

DeepSeek V3.2 (Reasoner) has a context window of 128K tokens, which means it can process up to 128,000 tokens in a single request.

Is DeepSeek V3.2 (Reasoner) open source?

Yes, DeepSeek V3.2 (Reasoner) is open source. The model weights are publicly available and can be self-hosted.

What is DeepSeek V3.2 (Reasoner) best used for?

DeepSeek V3.2 (Reasoner) is best suited for: reasoning, math, coding, research. o-series-class reasoning at a fraction of the price.

What is DeepSeek V3.2 (Reasoner)'s ELO score?

DeepSeek V3.2 (Reasoner) has an ELO score of 1385 on the LMSYS Chatbot Arena leaderboard, placing it among top-tier models.