Skip to main content

Apps & Servers Skills News Models Leaderboard Blog

DeepSeek-V3.2 — AI Model — MCP App Store | MCP App Store

Back to Deepseek V3

DeepSeek AI · Deepseek V3

DeepSeek-V3.2

instruct671BClosed source

DeepSeek's updated 671B MoE chat model — improved instruction-following and reasoning over V3, still at open-weight cost.

164K tokensMoEProprietary

Benchmarks

DeepSeek-V3.2

70.0%

DeepSeek V3.2 high reasoning — mini-SWE-agent v2

DeepSeek-V3.2

74.3%

WebTactix agent (Feb 2026)

DeepSeek-V3.2

56.7%

DeepSeek-V3.2-Exp

56.7%

DeepSeek-V3.2-Exp FC — rank #19

DeepSeek-V3.2

69.2 pass@1

DeepSeek-V3.2

42.9%

TTE-MatrixAgent + DeepSeek-V3.2 — rank #1, 42.9% resolved

DeepSeek-V3.2

25.1%

DeepSeek-V3.2

62.2%

DeepSeek-V3.2-Exp

58.9%

DeepSeek V3.2 Thinking — rank #23

DeepSeek-V3.2

40.8%

DeepSeek-V3.2

95.1%

DeepSeek-V3.2

90.2%

DeepSeek-V3.2

79.9%

DeepSeek-V3.2

78.3%

DeepSeek-V3.2

82.4%

DeepSeek-V3.2

39.3%

DeepSeek-V3.2

17.3%

DeepSeek-V3.2

51.4%

DeepSeek-V3.2

67.4%

DeepSeek-V3.2

62.2%

DeepSeek-V3.2

35.2%

DeepSeek-V3.2

1034.0 dollar

MCP App Store

The AI ecosystem directory — MCP Apps, Agent Skills, and daily news.

Directory

MCP Apps & Servers Agent Skills AI News AI Models What are MCP Apps?

Resources

Documentation Specification GitHub

Account

Sign in Get Started Dashboard

Company

Contact Build an MCP App

Legal

Privacy Policy Terms of Use

© 2026 MCP App Store

All rights reserved.

Other models in this family

DeepSeek Chat (V3.2)instruct671B

DeepSeek Reasoner (V3.2)reasoning671B

DeepSeek-V3instruct671B

DeepSeek-V3-0324instruct671B

DeepSeek-V3.1instruct671B

DeepSeek-V3.1-Basebase671B

DeepSeek-V3.2-Expinstruct671B