Atlas

Models

← All models

GPT-5.5

OpenAI · 2026-04-23

GPT-5.5 is OpenAI's frontier reasoning model for complex professional work, released as its smartest and most intuitive model to date. It accepts text and image inputs, supports configurable reasoning effort values of none, low, medium, high, and xhigh, and shows especially strong gains in agentic coding, computer use, knowledge work, and early scientific research.

Benchmark scores

BenchmarkScore
A-Fantasia34.7
A-Fantasia (Valid Attempts with Retries)34.7
A-Fantasia Backwards Spelling27.0
A-Fantasia Chess14.0
A-Fantasia Chess (Valid Attempts with Retries)14.0
A-Fantasia Cube (Valid Attempts with Retries)63.0
A-Fantasia Cube Rotation63.0
A-Fantasia Spelling (Valid Attempts with Retries)27.0
AA-AnalystAgent (pass@1)66.3
AA-AnalystAgent (pass@5)78.8
AA-AnalystAgent (pass^5)50.0
AA-Briefcase Elo1158
AA-Briefcase v1.1 Elo1136
AA-LCR74.3
AA-LCR v1.184.3
AA-Omniscience Index20.5
ActiveVision10.6
ActiveVision with tools37.6
AdvancedIF81.3
AdvancedIF (rubric mean score)95.3
Agent Red Teaming k=100 (CAIS)41.5
Agents' Last Exam (ALE)26.6
Agents' Last Exam ALE-CLI Pass Rate27.1
Agents' Last Exam ALE-CLI Score49.2
Agents' Last Exam Full-Spectrum Pass Rate23.6
Agents' Last Exam Full-Spectrum Score41.1
Agents' Last Exam Last-Exam Pass Rate2.6
Agents' Last Exam Last-Exam Score16.1
Agents' Last Exam Near-Term Pass Rate44.0
Agents' Last Exam Near-Term Score72.4
Loading Atlas data…