DeepSeek V3.2 Exp

Reasoning

Creator

DeepSeek

Released

2025-09-29

Intelligence

32.9

Artificial Analysis Index

Coding

33.3

Artificial Analysis Index

In $/1M

$0.28

input tokens

Out $/1M

$0.41

output tokens

Blended $/1M

$0.31

3:1 blended

Speed

0

tokens / sec

Benchmark breakdown

Independent evaluation scores, normalized to 0–100.

GPQA Diamond
80
Humanity’s Last Exam
14
MMLU-Pro
85
SciCode
38
LiveCodeBench
79
MATH-500
AIME 2025
88
τ²-Bench (agentic)
34
Terminal-Bench Hard
31
IFBench
54
Via Artificial Analysis