ModelsCompareBest forBenchmarksStatusPricingAPI

GPT-5.4

OpenAI·Operational·Released 2026-03-05textvision
Context1,050k
Max output128k
Input / 1M$2.5 ↗
Output / 1M$15 ↗
Cached in$0.25
Benchmarks21

Benchmarks

Cited public sources · rank is exact-benchmark, apples-to-apples

AIME 2026 (MathArena)#3/2199.2% ↗
FrontierMath Tier 4 v2 (Epoch AI run)#11/4449% ↗
FrontierMath Tiers 1-3 v2 (Epoch AI run)#11/5978.6% ↗
GPQA Diamond (Epoch AI run)#11/7793.3% ↗
OTIS Mock AIME 2024-2025 (Epoch AI run)#11/7697.8% ↗
SimpleQA Verified (Epoch AI run)#25/5345.1% ↗
SWE-bench Verified (Epoch AI run)#6/2476.9% ↗
LMArena Elo#15/621452.6 ↗
BrowseComp#6/1582.7% ↗
GDPval (wins or ties)#2/683% ↗
GPQA Diamond#11/7092.8% ↗
Humanity's Last Exam (no tools)#9/2939.8% ↗
Humanity's Last Exam (with tools)#8/1352.1% ↗
MCP Atlas#4/967.2% ↗
MMMU-Pro (no tools)#2/681.2% ↗
MMMU-Pro (with tools)#4/782.1% ↗
OSWorld-Verified#7/1275% ↗
SWE-Bench Pro#10/1557.7% ↗
Tau2-bench Telecom#2/798.9% ↗
Terminal-Bench 2.0#2/1375.1% ↗
Toolathlon#2/654.6% ↗

Compare

vs GPT-6 Astravs GPT-5.6 Solvs GPT-5.6 Terravs GPT-5.6 Lunavs Claude Fable 5.1vs Claude Opus 5vs Claude Sonnet 5vs Claude Haiku 4.5

More from OpenAI

GPT-6 AstraGPT-5.6 SolGPT-5.6 TerraGPT-5.6 LunaGPT-5.5GPT-5.5 Pro

Access GPT-5.4 and every other model through one endpoint with automatic failover: Respan gateway.