OpenAI Unveils GPT-5.4: A New Frontier in Professional Reasoning and Computer Control

OpenAI has officially pulled the curtain back on its latest powerhouse, GPT-5.4, marking a significant milestone in the evolution of artificial intelligence. This isn't just another incremental update; it is a model specifically engineered for high-level professional tasks. By integrating the newest breakthroughs in reasoning, coding, and complex workflows, GPT-5.4 builds upon the foundation of its predecessor, the GPT-5.3-Codex, to deliver a much more refined and capable assistant for the modern workforce.

Alongside the standard release, the company introduced GPT-5.4 Pro. This enhanced version is designed to be the heavy lifter of the family, capable of navigating intricate problems that would typically stump standard AI models. According to OpenAI, the core focus of this release is to make professional work—specifically within spreadsheets, presentations, and documents—far more efficient and intuitive than ever before.

The Power of Interactive Reasoning

One of the most striking features of this release is the GPT-5.4 Thinking model now available within ChatGPT. Unlike previous versions that simply provided an output, this model exposes its entire chain of reasoning. Users can watch the AI “think” through a problem before it arrives at a final answer. What makes this truly revolutionary is that it’s not a one-way street; users can actually adjust instructions mid-process, guiding the AI’s logic as it develops its response. This transparency ensures that the final result aligns perfectly with the user's specific intent.

Beyond just thinking better, GPT-5.4 is also much better at "seeing." The model can now recognize user interface elements from screenshots. This capability allows it to perform operations using a mouse and keyboard across various software applications, bridging the gap between a chatbot and a true digital employee that can interact with your computer much like a human would.

Breaking Records in Benchmarks and Reliability

When it comes to raw performance, GPT-5.4 is setting new records. In recent benchmark tests, the model either matched or outperformed seasoned industry professionals in a staggering 83 percent of GDPval comparisons. It also showed a marked improvement in coding capabilities, surpassing previous versions in its ability to write, debug, and optimize complex code.

Reliability is another major pillar of this update. OpenAI has worked hard to solve the issue of "hallucinations" or factual errors. GPT-5.4 boasts significantly higher factual accuracy and is far less likely to provide misleading information compared to the older GPT-5.2 model. This makes it a much more dependable tool for researchers, lawyers, and data analysts who require high-stakes accuracy.

Deployment and Cost Efficiency

For those eager to try it out, GPT-5.4 is currently available to users on ChatGPT Plus, Team, and Pro plans. Developers can also access the model via the API to integrate these advanced capabilities into their own applications. As OpenAI looks toward the future, they have announced plans to phase out GPT-5.2 Thinking over the next few months to streamline their offerings.

While the cost per individual token is slightly higher with this model, OpenAI notes that the overall efficiency of the tokens has improved. In many practical scenarios, this efficiency can actually lead to lower total costs for the user, proving that GPT-5.4 is not only smarter but also more economical in the long run. With better web search integration and a massive context window, GPT-5.4 is clearly positioned as the new standard for professional AI.