GPT-5.4: Controls Computers Now
Master your PC: GPT-5.4's native computer-use crushes benchmarks and automates workflows.
5 mar 2026 (Aggiornato il 18 mar 2026) - Scritto da Lorenzo Pellegrini
This image is part of OpenAI's official brand assets, available from their press kit
Lorenzo Pellegrini
5 mar 2026 (Aggiornato il 18 mar 2026)
OpenAI Releases GPT-5.4: First Model with Native Computer-Use Capabilities
OpenAI has launched GPT-5.4, marking a groundbreaking advancement in artificial intelligence with native computer-use capabilities that enable it to interact directly with operating systems and web environments. This release positions the model as the most capable frontier AI for professional tasks, addressing key limitations in reasoning, efficiency, and real-world application.
What is GPT-5.4 and Why Does It Matter?
GPT-5.4 builds on the GPT-5 series by introducing variants like GPT-5.4 Pro for high-performance needs and GPT-5.4 Thinking for advanced reasoning. Available immediately to ChatGPT Plus, Team, and Pro users, it replaces older versions such as GPT-5.2 Thinking. The model's native computer-use features allow it to excel in benchmarks like OSWorld-Verified and WebArena Verified, achieving record scores by autonomously navigating software interfaces and web tasks.
Key Features and Improvements
The standout upgrade is the massive context window of up to 1 million tokens in the API version, a significant leap from the 400,000 tokens in GPT-5.2. This expansion supports deeper data processing for coding, research, and complex workflows.
- Enhanced token efficiency solves problems with fewer tokens than predecessors.
- Record 83% score on GDPval benchmark for knowledge work tasks.
- 33% reduction in factual errors per claim and 18% fewer overall response errors compared to GPT-5.2.
- New Tool Search system optimizes API tool calling by dynamically retrieving definitions, reducing costs and speed in multi-tool setups.
Native Computer-Use Capabilities
GPT-5.4's computer-use prowess shines in verified benchmarks where it outperforms rivals by directly controlling computers. This includes tasks like software navigation and web interactions, making it ideal for automation in professional settings.
Context of the Rapid Release
The launch follows closely after GPT-5.3 Instant, teased just days prior amid reports of user migration to competitors like Claude. OpenAI's announcement of GPT-5.4 "sooner than you think" quelled concerns, especially following a controversial U.S. Department of Defense contract. While rivals paused similar deals over ethical issues, OpenAI proceeded with adjustments for defense applications.
Implications for Users and Developers
Professionals gain a powerful tool for efficient workflows, while developers benefit from improved API features. The model's reduced hallucinations and superior reasoning promise more reliable outputs across applications, from content creation to data analysis.
Conclusion
GPT-5.4 represents OpenAI's bold step toward practical, computer-integrated AI, redefining possibilities in automation and productivity.
In summary, with its pioneering computer-use abilities, expanded context, and benchmark dominance, GPT-5.4 sets a new standard. Enthusiasts and professionals should explore it now to leverage these transformative capabilities.
GPT-5.4 feels like the moment where “AI assistant” quietly turns into “AI coworker”: its native computer-use, huge context window, and better reasoning don’t just boost benchmarks, they change what kinds of problems you can safely hand off to a model, from multi-step research and complex workflows to full-on software and web navigation, while also raising fresh questions about how fast we should deploy systems that can independently act inside our digital environments
Which users have immediate access to GPT-5.4?