DeepSeek V3.1 Terminus

Reasoning

Creator

DeepSeek

Released

2025-09-22

Intelligence

33.9

Artificial Analysis Index

Coding

33.7

Artificial Analysis Index

In $/1M

$1.64

input tokens

Out $/1M

$2.75

output tokens

Blended $/1M

$1.91

3:1 blended

Speed

0

tokens / sec

Benchmark breakdown

Independent evaluation scores, normalized to 0–100.

GPQA Diamond
79
Humanity’s Last Exam
15
MMLU-Pro
85
SciCode
41
LiveCodeBench
80
MATH-500
AIME 2025
90
τ²-Bench (agentic)
37
Terminal-Bench Hard
30
IFBench
57
Via Artificial Analysis