DeepSeek V3.2 Exp Reasoning is DeepSeek's language model. A reasoning-mode experimental build of DeepSeek V3.2 (925 variant), combining sparse attention efficiency with chain-of-thought inference.
Specifications
Canonical IDdeepseek-v3-2-reasoning-925
TypeLanguage
StatusActive
CreatorDeepSeekDeepSeek
Input ModalitiesText
Output ModalitiesText
Benchmarks
Intelligence Index
32.9
#110
Coding Index
33.3
#83
Math Index
87.7
#39
MMLU-Pro
0.8
#30
GPQA
0.8
#92
HLE
0.1
#88
LiveCodeBench
0.8
#30
IFBench
0.5
#120
Time to First Token
SciCode
0.4
#122
AIME 2025
0.9
#39
LCR
0.7
#24
TerminalBench Hard
0.3
#78
TAU2
0.3
#203
Output TPS
0.0
#333

Capabilities

Input1/5
Text
Image·
Audio·
Video·
PDF·
Output1/5
Text
Image·
Audio·
Video·
Embedding·
Capabilities0/13
Reasoning·
Adaptive Reasoning·
Function Calling·
Parallel Function Calling·
Structured Outputs·
Native JSON Schema·
Web Search·
URL Context·
Computer Use·
Code Execution·
File Search·
Prompt Caching·
Assistant Prefill·

Versions

VersionReleasedContextInput / 1MOutput / 1MStatus
DeepSeek V3.2 Speciale164K$0.287$0.431Available
DeepSeek V3.2 Thinking128K$0.620$1.85Available
DeepSeek V3.2 Exp ReasoningCurrent
DeepSeek V3.2 ExpAvailable
DeepSeek V3.1 671B164KAvailable
DeepSeek V3 324BAvailable
DeepSeek V3 Turbo64K$0.400$1.30Available
DeepSeek V3.2 ReasoningAvailable
DeepSeek V3.1 Terminus ReasoningAvailable
DeepSeek V3.1 ReasoningAvailable
DeepSeek V3 2SP$0.580$1.68Available

Model IDs