Neura Intelligence Index

AI Model Benchmarks

Back to Leaderboard

DeepSeek-V3.2 (Thinking) vs GPT-5

DeepSeek-V3.2 (Thinking)

DeepSeek 🇨🇳

61.2

#134—

OpenAI 🇺🇸

62.2

#131—

Category Comparison

Benchmark-by-Benchmark

reasoning

Benchmark	DeepSeek-V3.2 (Thinking)	GPT-5
GPQA Diamond	82.4	85.7
Humanity's Last Exam	25.1	24.8

math

Benchmark	DeepSeek-V3.2 (Thinking)	GPT-5
AIME 2025	93.1	94.6
FrontierMath	—	26.3

coding

Benchmark	DeepSeek-V3.2 (Thinking)	GPT-5
SWE-bench Verified	73.1	74.9

agent

Benchmark	DeepSeek-V3.2 (Thinking)	GPT-5
BrowseComp	51.4	54.9
Toolathlon	35.2	—

multimodal

Benchmark	DeepSeek-V3.2 (Thinking)	GPT-5
MMMU	—	84.2
MMMU-Pro	—	78.4
CharXiv Reasoning	—	81.1

Category Winners

coding

GPT-5

71.1

math

DeepSeek-V3.2 (Thinking)

74.9

reasoning

GPT-5

57.9

multimodal

GPT-5

76.9

agent

GPT-5

29.8

DeepSeek-V3.2 (Thinking) vs GPT-5 — Benchmark Comparison | Neura Market