GPT‑5.4

Introducing GPT-5.4: The Frontier Model for Professional Work

OpenAI has officially released GPT-5.4, the most capable and efficient frontier model designed specifically for professional work. Available in ChatGPT (as GPT-5.4 Thinking), the API, and Codex, this model represents a significant leap in reasoning, coding, and agentic workflows. For users requiring maximum performance on the most complex tasks, GPT-5.4 Pro is also available, setting new benchmarks across industry-standard evaluations.

What's GPT-5.4?

GPT-5.4 is a general-purpose frontier model that brings together the latest advances in reasoning and tool use into a single, cohesive system. It incorporates the industry-leading coding capabilities of GPT-5.3-Codex while drastically improving how the model interacts with software environments, spreadsheets, and professional documents.

Unlike its predecessors, GPT-5.4 is designed to get complex real-world work done accurately and efficiently. It is the first mainline reasoning model to offer native, state-of-the-art computer-use capabilities, allowing AI agents to operate computers, navigate desktop environments, and execute multi-step workflows across various applications. Whether used through the GPT-5.4 Thinking interface or the GPT-5.4 Pro API, the model delivers higher-quality outputs with less "back and forth" required from the user.

Features of GPT-5.4

Advanced Professional Reasoning

GPT-5.4 excels in knowledge work across 44 occupations. On the GDPval benchmark, it matches or exceeds industry professionals in 83.0% of comparisons. It shows remarkable proficiency in:

• Spreadsheet Modeling: Achieving 87.3% on junior investment banking analyst tasks.
• Document Analysis: Reducing factual errors by 33% compared to GPT-5.2.
• Presentation Design: Human raters prefer GPT-5.4 presentations 68% of the time due to stronger aesthetics.

Native Computer Use and Vision

As the first model with native computer-use capabilities, GPT-5.4 can:

• Issue mouse and keyboard commands in response to screenshots.
• Navigate desktop environments via OSWorld-Verified with a 75.0% success rate.
• Support high-fidelity perception up to 10.24M total pixels for dense image understanding.

Coding and Developer Workflows

GPT-5.4 combines the strengths of GPT-5.3-Codex with faster token velocity.

• /fast mode in Codex delivers up to 1.5x faster token speeds.
• Playwright (Interactive): An experimental skill that allows the model to visually debug and test web apps as it builds them.
• 1M Token Context: Supports massive context windows for long-horizon task planning.

Agentic Tool Use and Tool Search

The introduction of tool search allows GPT-5.4 to work with ecosystems containing tens of thousands of tools. Instead of crowding the context window, the model searches for and appends tool definitions only when needed, reducing token usage by up to 47%.

Use Case Scenarios

• Financial Services: Use GPT-5.4 Pro to build complex financial models, legal analyses, and multi-page investment slide decks with high accuracy.
• Software Engineering: Developers can leverage the GPT-5.4 API to create agents that automate browser playtests, debug Electron apps, and manage bulk data entry across legacy portals.
• Legal and Compliance: Professional services can utilize GPT-5.4 for structuring complex transactional analyses and maintaining consistency across lengthy legal contracts.
• Project Management: In ChatGPT, GPT-5.4 Thinking provides an upfront plan, allowing users to adjust the model's course mid-response for better alignment with project goals.

How to Use GPT-5.4

In ChatGPT

Subscribers to ChatGPT Plus, Team, and Pro can access GPT-5.4 Thinking directly from the model picker. Enterprise and Edu users can enable early access via their admin settings. GPT-5.4 Pro is exclusively available to Pro and Enterprise plans for maximum reasoning effort.

In the API and Codex

• Standard Access: Use the model string gpt-5.4 for general tasks.
• High Performance: Use gpt-5.4-pro for the most demanding workloads.
• Long Context: Configure model_context_window to access the experimental 1M token context in Codex.
• Speed: Enable priority processing to access the fastest available token speeds for real-time applications.

FAQ

Q: How does GPT-5.4 compare to GPT-5.2 in terms of cost? A: While GPT-5.4 has a higher price per token ($2.50 / M input tokens), it is significantly more token-efficient. It uses fewer tokens to solve the same problems compared to GPT-5.2, often resulting in lower total costs for complex tasks.

Q: What is the difference between GPT-5.4 Thinking and GPT-5.4 Pro? A: GPT-5.4 Thinking is the standard reasoning model available in ChatGPT. GPT-5.4 Pro is designed for users who need maximum performance and the highest level of reasoning effort for extremely complex professional tasks.

Q: Is GPT-5.2 still available? A: Yes, GPT-5.2 Thinking will remain available as a Legacy Model for paid users until June 5, 2026.

Q: Can GPT-5.4 use tools like Excel? A: Yes. OpenAI has released a ChatGPT for Excel add-in specifically designed to leverage GPT-5.4 capabilities within spreadsheets.

Q: How does GPT-5.4 handle safety and cybersecurity? A: GPT-5.4 is deployed with "High" cyber capability protections under the Preparedness Framework. It includes an expanded safety stack with monitoring and asynchronous blocking to prevent misuse while reducing unnecessary refusals.