What is GPT‑5.4?

GPT‑5.4 is OpenAI’s most capable and efficient frontier model for professional work, released in March 2026. It’s available in ChatGPT (as GPT‑5.4 Thinking), the API, and Codex. The model combines advanced reasoning, industry-leading coding capabilities from GPT‑5.3‑Codex, and improved agentic workflows into a single system. GPT‑5.4 delivers deeper web research, stronger context retention on long tasks, and 33% fewer factual errors than its predecessor. A separate GPT‑5.4 Pro tier offers maximum performance for complex tasks.

Who it's for

• Knowledge professionals – Accountants, consultants, and analysts who need polished deliverables like financial models, slide decks, and spreadsheets with less back-and-forth.
• Software developers – Engineers using Codex or the API who want native computer-use capabilities, 1M-token context windows, and tool search for building reliable agents.
• Power researchers – Users who tackle highly specific queries requiring deep web research and sustained context across long, multi-step investigations.

Key features

Mid-response redirection

GPT‑5.4 Thinking can now provide an upfront plan of its reasoning, and you can interrupt the model mid-response to adjust course. Instead of starting over, you redirect the output while it’s working, arriving at a result more closely aligned with your needs in fewer turns.

Deep web research and context retention

The model improves deep web research for highly specific queries while better maintaining context for questions that require longer thinking. This means higher-quality answers that stay relevant to the task without drifting or losing track of earlier instructions.

Native computer-use capabilities

GPT‑5.4 is the first general-purpose model with state-of-the-art computer-use abilities, enabling agents to operate computers and carry out complex workflows across applications. It supports up to 1M tokens of context, allowing agents to plan, execute, and verify tasks across long horizons.

Token-efficient reasoning

GPT‑5.4 is OpenAI’s most token-efficient reasoning model yet, using significantly fewer tokens to solve problems compared to GPT‑5.2. This translates to reduced token usage and faster speeds, saving costs without sacrificing intelligence.

What stands out

“GPT‑5.4 delivers what you asked for with less back and forth—same intelligence, more control, less token burn by default.”

This isn’t just about raw benchmark scores. The model’s ability to be redirected mid-response, combined with its token efficiency, fundamentally changes how professionals interact with AI. On GDPval, which tests agents across 44 occupations, GPT‑5.4 matches or exceeds industry professionals in 83.0% of comparisons—a leap from GPT‑5.2’s 70.9%. It also achieves 82.7% on BrowseComp for web research and 75.0% on OSWorld-Verified for computer use, setting new state-of-the-art results across multiple dimensions.

Worth checking out if…

You’re a professional who needs reliable, polished outputs from complex, multi-step tasks—whether that’s building financial models, conducting deep research, or developing software agents. If you’ve been frustrated by models that lose context, burn tokens on unnecessary reasoning, or force you to start over when you want to adjust direction, GPT‑5.4 offers a more controlled and efficient workflow. It’s also worth exploring if you’re building agentic systems that need native computer-use capabilities and long-horizon planning.