DeepSeekDeepSeek V3.2Open SourceArena #41Dec 1, 2025

DeepSeek V3.2 Thinking

Thinking mode of DeepSeek V3.2

Context Window
128K
tokens
Max Output
64K
tokens
Released
Dec 1, 2025
Arena Rank
#41
of 305 models

Capabilities

👁Vision
🧠Reasoning
🔧Tool Calling
Prompt Caching
🖥Computer Use
🎨Image Generation

Pricing Comparison

RouterInput / 1MOutput / 1MCached Input / 1M
Vercel AI$0.28$0.42

Benchmarks

Arena ELOLMSYS Chatbot Arena
1,421/1,500

Model IDs

Tags

reasoningtool-useimplicit-caching
Compare with another model

Related Models