Robots Atlas>ROBOTS ATLAS
GPT-5.5 Pro
AI Models›GPT

GPT-5.5 Pro

gpt-5.5-pro · Family: GPT
GPT-5.5 Pro is a variant of GPT-5.5 applying additional parallel test-time compute on harder questions, targeting higher-accuracy work in business, legal, scientific, and data-intensive domains.
✓ Active⏳ Limited accessLLMMultimodalReasoning modelTool-using model📁 GPT
Context window
1M
tokens
Max output
128,000
tokens
Release date
23 April 2026
🏢OpenAIProducer
Access:APIHostedDeployment:☁ Cloud

Overview

GPT-5.5 Pro is a variant of OpenAI's GPT-5.5 model, announced on April 23, 2026. It applies additional parallel test-time compute on harder queries to deliver higher accuracy at the cost of longer response times.

The model supports text and image input with text output, a 1,050,000-token context window, and up to 128,000 output tokens. Its knowledge cutoff is December 1, 2025. It supports reasoning, function calling, structured outputs and hosted tools (web search, file search, code interpreter, image generation, hosted shell, MCP).

GPT-5.5 Pro is available in the OpenAI API via the Responses and Chat Completions endpoints, as well as Batch (50% discount) and Priority processing (2.5x standard rate). It does not support cached input discount, fine-tuning, streaming, audio or video. In ChatGPT, the model is available on the Pro, Business, and Enterprise plans. OpenAI classifies the cyber and biological capabilities of GPT-5.5-class models as "High" under its Preparedness Framework.

Classification
LLMMultimodalReasoning modelTool-using model
Family: GPT
Applications
CodingData analysisDocument analysisDocument generationResearch assistanceMultimodal document understanding
Access & deployment
APIHosted
Cloud
Weights: Closed
Key parameters
📏 Context: 1M
✓ Tools
📥 Input: text, image
Platforms
OpenAI API

Technical specification

Context window
1M
tokens
Max output tokens
128,000
tokens per response
Knowledge cutoff
1 Dec 2025
Knowledge boundary
License
Proprietary (closed)
Features:✓ Tool use
Modalities
⬇ Input
textimage
⬆ Output
textcodestructured_data

Capabilities and applications

Native model capabilities
Reasoning
The model's ability to reason logically and solve complex problems.
Category: reasoning
Multi-step reasoning
Carrying out multi-step chains of reasoning across long, complex tasks.
Category: reasoning
Long context
Maintaining coherence and focus across very long input context.
Category: language
Coding
Generating, analysing and modifying source code.
Category: coding
Function Calling
Category: planning
Structured output
Producing data in structured formats such as JSON.
Category: structured_generation
Audio understanding
Category: audio
Image understanding
Analysing and interpreting the content of images.
Category: vision
Video Understanding
Category: video
Chart understanding
Reading and interpreting charts, tables and diagrams.
Category: vision
Diagram reasoning
Category: reasoning
OCR
Recognising text within images and documents.
Category: vision
Multilingual
Understanding and generating text in many languages.
Category: language
Planning
Forming and executing action plans for complex tasks.
Category: planning
Streaming output
Category: reasoning
Interleaved Multimodal Input
Category: reasoning
Multimodal understanding
Category: multimodal
Application domains
CodingData analysisDocument analysisDocument generationResearch assistanceMultimodal document understanding

Benchmark results

4 benchmarks
BrowseComp
accuracy · Evaluation of the ability to retrieve difficult information from the web; reasoning effort xhigh in a research environment.
90.1%
📅 23 Apr 2026📄 OpenAI (openai.com/index/introducing-gpt-5-5/)
GPT-5.5 Pro scored 90.1%, above GPT-5.5 standard (84.4%) and Gemini 3.1 Pro (85.9%). Evaluated with reasoning effort xhigh.
EpochAI Frontier Math
accuracy · Evaluation with xhigh reasoning effort in a research environment.
52.4%
📅 23 Apr 2026📄 OpenAI (openai.com/index/introducing-gpt-5-5/)
GPT-5.5 Pro scored 52.4%, slightly above GPT-5.5 standard at 51.7%.
EpochAI Frontier Math
accuracy · Hardest FrontierMath tier; reasoning effort xhigh in a research environment.
39.6%
📅 23 Apr 2026📄 OpenAI (openai.com/index/introducing-gpt-5-5/)
GPT-5.5 Pro scored 39.6%, above GPT-5.5 standard (35.4%) and Claude Opus 4.7 (22.9%).
Humanity's Last Exam (HLE)
accuracy · Evaluation with tools enabled; reasoning effort set to xhigh in a research environment.
57.2%
📅 23 Apr 2026📄 OpenAI (openai.com/index/introducing-gpt-5-5/)
GPT-5.5 Pro scored 57.2%, above GPT-5.5 standard (52.2%) and Claude Opus 4.7 (54.7%).

Pricing

Technical architecture

Deployment and security

☁ Available on platforms
☁OpenAI APIPlatform
🔒 Security / Enterprise
✓ Verified enterprise information

OpenAI rates the cyber and biological capabilities of GPT-5.5 models (including Pro) as High under its Preparedness Framework. The model underwent a full pre-release safety process, including targeted evaluations for advanced cyber and biological capabilities.

Updated: 25 Apr 2026↗ Security documentation

Sources and related pages