DeepSeek V3.1

Reasoning

Creator

DeepSeek

Released

2025-08-21

Intelligence

27.7

Artificial Analysis Index

Coding

29.7

Artificial Analysis Index

In $/1M

$0.59

input tokens

Out $/1M

$1.69

output tokens

Blended $/1M

$0.86

3:1 blended

Speed

0

tokens / sec

Benchmark breakdown

Independent evaluation scores, normalized to 0–100.

GPQA Diamond
78
Humanity’s Last Exam
13
MMLU-Pro
85
SciCode
39
LiveCodeBench
78
MATH-500
AIME 2025
90
τ²-Bench (agentic)
37
Terminal-Bench Hard
25
IFBench
41
Via Artificial Analysis