OpenAI Launches GPT-5.4 with Advanced Reasoning and Computer-Use Capabilities

Get ready, AI enthusiasts! OpenAI just dropped a major update on March 5, 2026, with the release of GPT-5.4 and GPT-5.4 Pro. Hailed as their most capable and efficient frontier model to date, GPT-5.4 is designed to supercharge professional work, blending advanced reasoning, coding, and truly agentic workflows. For those seeking maximum performance on the most complex tasks, GPT-5.4 Pro is also available. You can dive deeper into the specifics of this new launch by checking out the OpenAI Product Releases News.

What it's for

GPT-5.4 is built for serious professional work, offering a seamless integration of cutting-edge AI capabilities. It takes the industry-leading coding prowess of GPT-5.3-Codex and enhances it with robust tool usage across software environments and tasks involving spreadsheets, presentations, and documents.

If you're using ChatGPT, the new GPT-5.4 Thinking feature allows the model to present an upfront plan of its thought process. This means you can adjust its direction mid-response, ensuring the final output aligns perfectly with your needs. It also brings improved deep web research for highly specific queries and maintains context better for longer, more complex questions.

For developers and agents interacting via the API or Codex, GPT-5.4 is a game-changer. It's the first general-purpose model with native, state-of-the-art computer-use capabilities, enabling agents to operate computers and execute complex workflows across applications. With support for up to 1 million tokens of context, agents can plan, execute, and verify tasks over long horizons. Plus, it's OpenAI's most token-efficient reasoning model yet, significantly reducing token usage and increasing speed compared to GPT-5.2.

Why it matters

This isn't just another incremental update. GPT-5.4 delivers a leap forward in AI performance and reliability.

Unmatched Performance: GPT-5.4 sets new benchmarks across the board. On the GDPval Benchmark, which assesses knowledge work, GPT-5.4 matches or exceeds industry professionals in a remarkable 83.0% of comparisons, up from 70.9% for GPT-5.2. It also crushed OSWorld-Verified with a 75.0% success rate, surpassing human performance at 72.4%.
Reduced Hallucinations: Say goodbye to frustrating errors! GPT-5.4 is significantly more factual, with individual claims 33% less likely to be false and full responses 18% less likely to contain any errors compared to GPT-5.2.
Computer Mastery: This model excels at writing code to operate computers, whether via libraries like Playwright or by issuing mouse and keyboard commands from screenshots. It achieved a 67.3% success rate on WebArena-Verified, showcasing its real-world computer interaction skills.
Enhanced Coding: Combining the strengths of GPT-5.3-Codex with new knowledge and computer-use capabilities, GPT-5.4 matches or outperforms its predecessor on SWE-Bench Pro, offering lower latency and faster token velocity in Codex's /fast mode.
Smarter Tool Use: GPT-5.4 introduces tool search in the API, dramatically reducing token usage for tool-heavy workflows by allowing the model to look up tool definitions only when needed.

These advancements mean more reliable agents, faster developer workflows, and higher-quality outputs across the entire OpenAI ecosystem. For more in-depth technical details, you can always consult the OpenAI Research Releases Index.

Where you get it

GPT-5.4 is widely available for you to start leveraging its capabilities:

ChatGPT: Access GPT-5.4 as "GPT-5.4 Thinking" in ChatGPT, with GPT-5.4 Pro also available for peak performance.
API: Developers can integrate GPT-5.4 and GPT-5.4 Pro into their applications via the OpenAI API.
Codex: Utilize GPT-5.4 within Codex for enhanced coding and computer-use tasks.

Read more: Introducing GPT-5.4 for details on how to access these new frontier models and revolutionize your professional workflow.