GPT‑5.4
LatestOpenAI
•
Proprietary# 8
Released
Mar 5, 2026
# 6
Context Length
272K
Benchmarks
# 6
Code RankedAGI
83.2%
# 4
SWEBench Pro
57.7%
# 3
Terminal Bench 2.0
75.1%
# 5
Agentic RankedAGI
81.2%
# 3
BrowseComp
82.7%
# 7
Code Arena
1457
# 7
Code Livebench
77.5%
# 1
AgenticCode LiveBench
70.0%
# 3
OSWorld Verified
75.0%
# 4
Svelte Bench
95.6%
# 1
Reason RankedAGI
82.4%
# 9
HLE
39.8%
# 8
HLE w/ Tools
52.1%
# 5
GPQA Diamond
92.8%
# 4
Text Arena
1482
# 4
AIME 2026
95.2%
# 4
Vending Bench 2
$6144.18
# 2
NYT Connections
94.0%
# 2
MMMU Pro
81.2%
# 1
MMMU Pro w/ Tools
81.5%
# 17
Math RankedAGI
68.0%
# 2
RAGI Overall
70.7%
# 3
ARC AGI 2.0
76.1%
# 1
LiveCodeBench Pro
87.5%
# 3
𝜏²-Bench Telecom
98.9%
# 3
HealthBench Hard
40.1%
# 2
MedXpertQA Text
59.6%
# 3
MedXpertQA MM
77.1%
# 3
DeepSearch QA
73.6%
# 1
GDPval AA
1672
# 1
ZeroBench
41.0%
# 1
MCP Atlas
67.2%
Pricing
# 32
Input Cost /M
$2.5
# 38
Output Cost /M
$15
# 14
Cached Cost /M
$0.25