| AI models ranked by latest benchmarks

GPT‑5.4

Latest

OpenAI

•

Proprietary

# 20

Released

Mar 5, 2026

# 3

Knowledge Cutoff

Aug 25

# 3

Context Length

1M

Benchmarks

# 11

Code RankedAGI

81.2%

# 7

SWEBench Pro

57.7%

# 4

Terminal Bench 2.0

75.1%

# 14

Agentic RankedAGI

75.6%

# 9

Browse Comp

82.7%

# 14

Code Arena

1457

# 10

Code Livebench

77.5%

# 2

DeepSWE

56.0%

# 1

AgenticCode LiveBench

70.0%

# 6

OSWorld Verified

75.0%

# 5

Svelte Bench

95.6%

# 3

Cyber Gym

79.0%

# 3

Codeforces ELO

3168

# 11

Reason RankedAGI

77.6%

# 13

HLE

39.8%

# 13

HLE w/ Tools

52.1%

# 7

GPQA Diamond

92.8%

# 4

Text Arena

1482

# 5

AIME 2026

95.2%

# 6

Vending Bench 2

$6,144.18

# 2

NYT Connections

94.0%

# 6

MMLU Pro

87.5%

# 3

MMMU Pro

81.2%

# 2

MMMU Pro w/ Tools

82.1%

# 6

Math RankedAGI

76.5%

# 9

RAGI RankedAGI

69.9%

# 7

ARC AGI 2.0

73.3%

# 1

LiveCodeBench Pro

87.5%

# 13

𝜏²-Bench Telecom

92.8%

# 3

HealthBench Hard

40.1%

# 2

MedXpertQA Text

59.6%

# 3

MedXpertQA MM

77.1%

# 6

DeepSearch QA

73.6%

# 4

GDPval AA

1674

# 1

ZeroBench

41.0%

# 7

MCP Atlas

70.6%

# 8

Finance Agent

56.0%

# 4

CharXiv Reasoning

82.8%

# 16

Code DesignArena

1282

# 3

Toolathlon Pass@1

54.6%

# 1

𝜏³ Bench

72.9%

Pricing

# 38

Input Cost /M

$2.5

# 45

Output Cost /M

$15

# 20

Cached Cost /M

$0.25

Additional Details

Model ID

gpt-5.4

Links

Last Updated

2 days ago