DeepSeek R1

Jan '25

Creator

DeepSeek

Released

2025-01-20

Intelligence

18.8

Artificial Analysis Index

Coding

15.9

Artificial Analysis Index

In $/1M

$1.68

input tokens

Out $/1M

$4.70

output tokens

Blended $/1M

$2.43

3:1 blended

Speed

0

tokens / sec

Benchmark breakdown

Independent evaluation scores, normalized to 0–100.

GPQA Diamond
71
Humanity’s Last Exam
9
MMLU-Pro
84
SciCode
36
LiveCodeBench
62
MATH-500
97
AIME 2025
68
τ²-Bench (agentic)
11
Terminal-Bench Hard
6
IFBench
39
Via Artificial Analysis