Robots Atlas>ROBOTS ATLAS
GPT-5.4 Thinking
AI Modelsโ€บGPT

GPT-5.4 Thinking

5.4 Thinkingย ยทย Family: GPT
The most advanced GPT-5 model for complex professional work, reasoning, coding, tool use, and long-context tasks.
Context window
1,050,000 tokens
tokens
Release date
5 March 2026

Overview

GPT-5.4 Thinking is an extension of the earlier GPT-5 models, oriented toward more complex professional work. The key changes relative to previous versions include improved handling of agentic workflows, noticeably enhanced computer use and tool search, more reliable context retention during long reasoning chains, and greater token efficiency compared to GPT-5.2. OpenAI also highlights improvements in deep web research, particularly for highly specific queries. In ChatGPT, the Thinking variant adds the ability to display a reasoning plan upfront, allowing the user to redirect the response before it is completed.

Classification
Family: GPT
Key parameters
๐Ÿ“ Context: 1,050,000 tokens
โœ“ Tools
๐Ÿ“ฅ Input: text, image, documents

Technical specification

Context window
1,050,000 tokens
tokens
License
Proprietary (commercial; available via OpenAI services under Terms of Use / Services Agreement)
Hardware requirements
No hardware requirements have been publicly disclosed. The model operates as an OpenAI cloud service (ChatGPT / API / Codex) and is not intended to run locally on user hardware.
Features:โœ“ Tool use
Modalities
โฌ‡ Input
textimagedocuments
โฌ† Output
codereportsstructured_datasummariestext

Capabilities and applications

Native model capabilities
Reasoning
Category: reasoning
Multi-step reasoning
Category: reasoning
Long context
Category: reasoning
Coding
Category: coding
Function Calling
Category: planning
Structured output
Category: structured_generation
Image understanding
Category: vision
OCR
Category: vision
Multilingual
Category: language
Planning
Category: planning
Streaming output
Category: reasoning

Deployment and security

๐Ÿ”’ Security / Enterprise
โœ“ Verified enterprise information

GPT-5.4 Thinking inherits the publicly documented platform-level security measures of OpenAI. Enterprise and security information is documented primarily for the API, ChatGPT Enterprise, Business, and Edu tiers, rather than as a separate security profile for the Thinking variant.

It should be noted in the catalog that the security information reflects platform-level inherited security, not benchmark security of a specific model variant.
Updated: 15 Mar 2026โ†— Security documentation