NEWHow AI models draw the world map · Gecko Tests
Beta
Home/Models/DeepSeek V3.2
DeepSeek logo

DeepSeek V3.2

by DeepSeek · Released Dec 2025

Open Source
51.5
avg score
Rank #145
Compare
Better than 56% of all models
Context
164K tokens (~82 books)
Input $/1M
$0.26
Output $/1M
$0.38
Type
text
License
Open Source
Benchmarks
35 tested
Data as of
About

DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...

Tested on 35 benchmarks · BenchGecko score 51.5. Top scores: Chatbot Arena Elo · Overall (1424.8%), Chatbot Arena Elo · Coding (1324.8%), OpenCompass · AIME2025 (93.0%).

Looking for similar performance at lower cost?
Gemma 2 9B scores 51.1 (99% as good) at $0.03/1M input · 88% cheaper
Capabilities
coding
62.3
#36 globally
reasoning
37.6
#122 globally
math
57.4
#110 globally
knowledge
51.9
#123 globally
agentic
21.3
#62 globally
speed
43.8
#52 globally
general
27.5
#151 globally
language
59.0
#106 globally
Benchmark Scores
Compare All
Tested on 35 benchmarks · Ranked across 9 categories
Score Distribution (all 326 models)
0255075100
▲ You are here
codingCompare coding →
LiveBench · Coding

Regularly refreshed coding problems that avoid data contamination. New problems added monthly to prevent memorization.

75.7·
OpenCompass · LiveCodeBenchV6

OpenCompass Live Code Bench v6. Fresh competitive programming problems to evaluate code generation without memorization.

75.4·
Aider polyglot

Multi-language code editing from Aider. Tests editing ability across Python, JavaScript, TypeScript, Java, C++, Go, Rust, and more.

74.2·
reasoningCompare reasoning →
ARC-AGI

Abstraction and Reasoning Corpus. Tests fluid intelligence through novel visual pattern recognition puzzles. Core measure of general intelligence.

57.0·
LiveBench · Data Analysis

Fresh data analysis tasks testing ability to interpret tables, charts, and statistical data.

45.0·
LiveBench · Reasoning

Regularly refreshed reasoning problems testing logical deduction, spatial reasoning, and analytical thinking.

44.3·
mathCompare math →
OpenCompass · AIME2025

OpenCompass evaluation on AIME 2025 problems. Tests mathematical reasoning on fresh competition problems.

93.0·
OTIS Mock AIME 2024-2025

Mock AIME (American Invitational Mathematics Exam) problems from OTIS. Tests mathematical competition performance.

87.8·
LiveBench · Mathematics

Regularly updated math problems that test numerical reasoning, algebra, calculus, and combinatorics.

64.0·
Quick compare:minimax logovs MiniMax M2.5DeepSeek logovs DeepSeek V3.2 ExpDeepSeek logovs DeepSeek V3.1
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Model Family · DeepSeek DeepSeek V3.2
Dec 2025
51.5
$0.26/M in164Kctx35 benchmarks
Sep 2025
51.6+0.1
$0.27/M in(+0.01)164Kctx18 benchmarks
Dec 2025
95.2+43.6
$0.40/M in(+0.13)164Kctx9 benchmarks
See the full DeepSeek V3.2 family →
Recently Happened
DeepSeek V3.2 pricing increased 5%
Oct 1, 2026
DeepSeek V3.2 pricing increased 5%
Sep 1, 2026
DeepSeek V3.2 pricing dropped 5%
Aug 22, 2026
DeepSeek V3.2 pricing dropped 6%
Jul 10, 2026
DeepSeek V3.2 pricing dropped 10%
Apr 21, 2026
DeepSeek V3.2 pricing increased 11%
Apr 18, 2026
Similar Models
minimax logo
MiniMax M2.5
minimax
51.4$0.30/1M
DeepSeek logo
DeepSeek V3.2 Exp
DeepSeek
51.6$0.27/1M
DeepSeek logo
DeepSeek V3.1
DeepSeek
51.8$0.25/1M
Links
Info
DeepSeekPricing explorerDevelopers · API
Research
Technical Report
Documentation
API DocsPlayground
Community
@DeepSeek
Source Code
GitHubHugging Face
BenchGecko API
deepseek-v3-2
Specifications
  • Typetext
  • Context164K tokens (~82 books)
  • ReleasedDec 2025
  • LicenseOpen Source
  • StatusActive
  • Cost / Message~$0.001
Available On
DeepSeek logoDeepSeek$0.26
Categories
codingreasoningmathknowledgeagenticspeedgenerallanguage
Learn More
context-windowtransformeropen weightstokens
Share & Export
Tweet
Related Models
minimax logoMiniMax M2.5
51.4
DeepSeek logoDeepSeek V3.2 Exp
51.6
DeepSeek logoDeepSeek V3.1
51.8
Google DeepMind logoGemma 2 9B
51.1
DeepSeek V3.2 is an open-source text AI model by DeepSeek, released in December 2025. It has an average benchmark score of 51.5. Context window: 164K tokens.

Related Models

MiniMax M2.5 · minimaxDeepSeek V3.2 Exp · DeepSeekDeepSeek V3.1 · DeepSeekGemma 2 9B · Google DeepMindPhi 3.5 Mini Instruct · Microsoft

Benchmarks

Chatbot Arena Elo · OverallChatbot Arena Elo · CodingOpenCompass · AIME2025OpenCompass · IFEvalOTIS Mock AIME 2024-2025

Related Pages

DeepSeek · ProviderDeepSeek · EconomyAll ModelsCompare ModelsPricingDevelopers · APIContext Window · Glossary

Key facts · as of 2026-11-11

  • DeepSeek V3.2 by DeepSeek. BenchGecko score 51.5, rank 145 of 326 scored models (normalized average of public benchmark scores).
  • List price $0.26 input · $0.38 output per 1M tokens (as of 2026-11-11).
  • Sold by 6 providers (as of 2026-11-11): GMICloud (fp8) $0.21 in / $0.31 out · SiliconFlow (fp8) $0.26 in / $0.80 out · AtlasCloud (fp8) $0.26 in / $0.38 out · DeepInfra (fp4) $0.26 in / $0.38 out · Baidu (fp8) $0.28 in / $0.42 out · and 1 more. Every provider

How to cite · data as of 2026-11-11

DeepSeek V3.2 · benchmarks, pricing and providers. BenchGecko, data as of 2026-11-11. https://benchgecko.ai/model/deepseek-v3-2

Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP