Home/Models/DeepSeek V3.2 Speciale
DeepSeek logo

DeepSeek V3.2 Speciale

by DeepSeek · Released Dec 2025

Open Source
95.2
avg score
Rank #4
Compare
Top 2% of all models
Context
164K tokens (~82 books)
Input $/1M
$0.40
Output $/1M
$1.20
Type
text
License
Open Source
Benchmarks
9 tested
Data updated today
About

DeepSeek-V3.2-Speciale is a high-compute variant of DeepSeek-V3.2 optimized for maximum reasoning and agentic performance. It builds on DeepSeek Sparse Attention (DSA) for efficient long-context processing, then scales post-training reinforcement learning...

Tested on 9 benchmarks with 78.2% average. Top scores: OpenCompass — AIME2025 (96.0%), OpenCompass — IFEval (91.7%), OpenCompass — GPQA-Diamond (86.7%).

Looking for similar performance at lower cost?
Qwen3.5 397B A17B scores 96.3 (101% as good) at $0.39/1M input · 3% cheaper
Capabilities
coding
80.9
#6 globally
math
96.0
#2 globally
knowledge
66.9
#25 globally
language
91.7
#7 globally
speed
37.4
#46 globally
Benchmark Scores
Compare All
Tested on 9 benchmarks · Ranked across 5 categories
Score Distribution (all 233 models)
0255075100
▲ You are here
OpenCompass — LiveCodeBenchV6

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

80.9
OpenCompass — AIME2025

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

96.0
OpenCompass — GPQA-Diamond

OpenCompass evaluation of GPQA Diamond. PhD-level science questions from the hardest subset.

86.7
OpenCompass — MMLU-Pro

OpenCompass MMLU-Pro evaluation. Harder knowledge test with more answer choices.

85.5
OpenCompass — HLE

OpenCompass evaluation of Humanitys Last Exam. Expert-level cross-discipline knowledge test.

28.6
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Recently Happened
DeepSeek V3.2 Speciale released
Mar 22, 2026
Links
Documentation
Community
BenchGecko API
deepseek-v3-2-speciale
Specifications
  • Typetext
  • Context164K tokens (~82 books)
  • ReleasedDec 2025
  • LicenseOpen Source
  • StatusActive
  • Cost / Message~$0.002
Available On
DeepSeek logoDeepSeek$0.40
Share & Export
Tweet
DeepSeek V3.2 Speciale is an open-source text AI model by DeepSeek, released in December 2025. It has an average benchmark score of 95.2. Context window: 164K tokens.