Home/Models/DeepSeek V3.2 Exp
DeepSeek logo

DeepSeek V3.2 Exp

by DeepSeek · Released Sep 2025

Open Source
58.7
avg score
Rank #73
Compare
Better than 69% of all models
Context
164K tokens (~82 books)
Input $/1M
$0.27
Output $/1M
$0.41
Type
text
License
Open Source
Benchmarks
14 tested
Data updated today
About

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...

Tested on 14 benchmarks with 53.2% average. Top scores: Chatbot Arena Elo — Overall (1422.8%), Chatbot Arena Elo — Coding (1285.5%), Fiction.LiveBench (83.3%).

Looking for similar performance at lower cost?
DeepSeek V3.2 scores 58.7 (100% as good) at $0.25/1M input · 7% cheaper
Capabilities
coding
55.9
#50 globally
reasoning
44.9
#56 globally
math
64.4
#41 globally
knowledge
66.6
#28 globally
agentic
42.9
#5 globally
language
42.5
#111 globally
Benchmark Scores
Compare All
Tested on 14 benchmarks · Ranked across 7 categories
Score Distribution (all 233 models)
0255075100
▲ You are here
Aider polyglot

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

74.2
LiveBench — Coding

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

73.2
WeirdML

Unusual and adversarial machine learning challenges. Tests robustness of reasoning about edge cases in ML systems.

39.5
LiveBench — Reasoning

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

45.5
LiveBench — Data Analysis

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

44.3
LiveBench — Mathematics

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

64.4
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Links
Documentation
Community
BenchGecko API
deepseek-v3-2-exp
Specifications
  • Typetext
  • Context164K tokens (~82 books)
  • ReleasedSep 2025
  • LicenseOpen Source
  • StatusActive
  • Cost / Message~$0.001
Available On
DeepSeek logoDeepSeek$0.27
Share & Export
Tweet
DeepSeek V3.2 Exp is an open-source text AI model by DeepSeek, released in September 2025. It has an average benchmark score of 58.7. Context window: 164K tokens.