OpenAI has officially introduced GPT-5.4, the newest upgrade in its GPT-5 series of artificial intelligence models. The update brings improvements in reasoning, coding performance, and workflow automation. Alongside the main model, the company has also launched a more powerful version called GPT-5.4 Pro.
One of the most notable additions in this release is the ability for the AI system to interact with computer environments, enabling more advanced automation and task execution.
Also read: OpenAI Signals Upcoming GPT-5.4 Model Soon After Launching GPT-5.3 Instant
Availability Across ChatGPT and Developer Platforms
The new GPT-5.4 model is available both for ChatGPT users and developers through OpenAI’s API platform. In ChatGPT, the model appears as GPT-5.4 Thinking, replacing the previous GPT-5.2 Thinking model for certain subscription tiers.
At launch, access to GPT-5.4 is limited to:
- ChatGPT Plus subscribers
- Team users
- Pro subscribers
Meanwhile, the GPT-5.4 Pro version is available only to Pro and Enterprise users, reflecting its focus on more demanding professional and business applications.
Improved Reasoning and Coding Performance
OpenAI says GPT-5.4 introduces improvements in several areas, particularly reasoning ability and programming support. The model builds on earlier updates by incorporating coding capabilities from previous systems while enhancing general-purpose performance.
With these upgrades, GPT-5.4 can assist with tasks such as:
- Writing and debugging software code
- Handling structured data in spreadsheets
- Generating presentations and reports
- Managing multi-step workflows involving different tools
These improvements aim to make the model more useful for developers, businesses, and advanced productivity tasks.
AI That Can Interact With Computers
A key highlight of GPT-5.4 is its computer-use capability. This feature allows the AI to operate inside a controlled computer environment and perform actions similar to a human user.
Using this capability, the model can:
- Interact with software applications
- Navigate websites and online tools
- Execute commands through keyboard and mouse inputs
- Automate complex digital workflows
Instead of only providing instructions, the AI can generate and execute commands that control a computer system through an automated process.
This type of functionality opens the door to more advanced AI agents that can complete tasks independently within digital environments.
Performance Benchmarks
OpenAI also shared performance results from internal testing to highlight improvements in the new model.
On the OSWorld-Verified benchmark, which evaluates how well AI models interact with desktop environments, GPT-5.4 achieved a success rate of approximately 75 percent.
The model also performed strongly on visual reasoning evaluations. On the MMMU-Pro benchmark, which measures an AI system’s ability to interpret visual information, GPT-5.4 achieved a score of around 81 percent.
These results indicate that the model has improved visual perception and better understanding of graphical interfaces.
Additional Safety Measures
Because GPT-5.4 includes powerful automation capabilities, OpenAI has introduced additional safety protections to monitor and control how the model operates.
The company has classified the model as having high cyber capability, which means stronger safeguards are required when deploying it.
Some of the safety measures include:
- Monitoring systems that track AI behavior
- Access restrictions in sensitive environments
- Mechanisms that block unsafe operations
- Evaluation tools designed to analyze how the model explains its reasoning
These protections aim to ensure that the model can be used responsibly while minimizing potential risks.
The Evolution of AI Assistants
The release of GPT-5.4 highlights a growing shift in how artificial intelligence is being developed. Modern AI systems are moving beyond simple chat-based interactions and are becoming capable of performing tasks directly within digital environments.
By combining reasoning, coding, and computer interaction abilities, AI assistants are gradually evolving into tools that can automate real work rather than simply provide information.
Also read: Google Unveils Gemini 3.1 Flash-Lite, a Faster and More Efficient AI Model for Developers
Final Thoughts
The launch of GPT-5.4 and GPT-5.4 Pro represents another step forward in OpenAI’s efforts to expand the capabilities of its AI models. With improvements in reasoning, coding support, and the ability to interact with computer systems, the new models aim to handle more advanced tasks.
Although access is currently limited to certain subscription tiers and enterprise users, GPT-5.4 demonstrates how AI technology is progressing toward more powerful and practical digital assistants.