Neura Intelligence Index

AI Model Benchmarks

Back to Leaderboard

DeepSeek-V4-Pro-Max vs GPT-5.2

DeepSeek-V4-Pro-Max

DeepSeek 🇨🇳

73.2

#51—

OpenAI 🇺🇸

73.3

#48—

Category Comparison

Benchmark-by-Benchmark

reasoning

Benchmark	DeepSeek-V4-Pro-Max	GPT-5.2
GPQA Diamond	90.1	92.4
Humanity's Last Exam	48.2	34.5
ARC-AGI v2	—	52.9

coding

Benchmark	DeepSeek-V4-Pro-Max	GPT-5.2
SWE-bench Verified	80.6	80
SWE-Bench Pro	55.4	—

general

Benchmark	DeepSeek-V4-Pro-Max	GPT-5.2
SimpleQA	57.9	—
MMMLU	—	89.6

agent

Benchmark	DeepSeek-V4-Pro-Max	GPT-5.2
BrowseComp	83.4	65.8
Toolathlon	51.8	46.3
MCP Atlas	73.6	60.6

math

Benchmark	DeepSeek-V4-Pro-Max	GPT-5.2
AIME 2025	—	100
FrontierMath	—	40.3

multimodal

Benchmark	DeepSeek-V4-Pro-Max	GPT-5.2
MMMU-Pro	—	79.5
CharXiv Reasoning	—	82.1
ScreenSpot Pro	—	86.3

Category Winners

coding

GPT-5.2

81.1

math

GPT-5.2

71.3

reasoning

DeepSeek-V4-Pro-Max

86.0

general

DeepSeek-V4-Pro-Max

77.7

multimodal

GPT-5.2

83.1

agent

DeepSeek-V4-Pro-Max

68.5

DeepSeek-V4-Pro-Max vs GPT-5.2 — Benchmark Comparison | Neura Market