Atlas

Models

← All models

GPT-5.4

OpenAI · 2026-03-05

GPT-5.4 is OpenAI's frontier model for complex professional work, combining advanced reasoning, coding, and agentic workflows with configurable reasoning effort from none through xhigh. It is OpenAI's first general-purpose model with native computer-use capabilities, incorporates the coding capabilities of GPT-5.3-Codex, and reduces factual errors relative to GPT-5.2.

Benchmark scores

BenchmarkScore
A-Fantasia42.3
A-Fantasia Backwards Spelling26.0
A-Fantasia Chess33.0
A-Fantasia Cube Rotation68.0
AA-LCR74.0
AA-Omniscience Index5.7
Agent Red Teaming k=100 (CAIS)59.7
Agents' Last Exam (ALE)20.5
Agents' Last Exam ALE-CLI Pass Rate22.5
Agents' Last Exam ALE-CLI Score41.8
Agents' Last Exam Full-Spectrum Pass Rate19.4
Agents' Last Exam Full-Spectrum Score34.3
Agents' Last Exam Last-Exam Pass Rate0.0
Agents' Last Exam Last-Exam Score8.1
Agents' Last Exam Near-Term Pass Rate33.6
Agents' Last Exam Near-Term Score57.8
Agents' Last Exam Overall Score37.3
Agents' Last Exam Unlicensed Full-Spectrum Pass Rate19.3
Agents' Last Exam Unlicensed Full-Spectrum Score34.9
Agents' Last Exam Unlicensed Last-Exam Pass Rate0.0
Agents' Last Exam Unlicensed Last-Exam Score10.3
Agents' Last Exam Unlicensed Near-Term Pass Rate33.6
Agents' Last Exam Unlicensed Near-Term Score57.8
Agents' Last Exam Unlicensed Overall Pass Rate22.1
Agents' Last Exam Unlicensed Overall Score39.9
AIME 2024 (avg@8)97.1
AIME 2024-202596.7
AIME 2025 (avg@8)96.3
AIME 202699.2
ALE-Bench1811
Loading Atlas data…