OpenAI launched GPT-5.4, a new frontier model designed for professional workloads that integrates advanced reasoning, encryption, and agency workflows into a single system.
The model is distributed across ChatGPT, API and Codex, and a high-quality GPT-5.4 Pro version is also available for complex tasks that require maximum computation and accuracy.
GPT-5.4 integrates the encryption capabilities introduced in the GPT-5.3 Codex while improving performance in tools, software environments, and professional tasks such as spreadsheets, presentations, and document creation. The model can also pre-determine the outline of their reflections within ChatGPT, allowing users to guide responses in the medium.
The publication introduces native computing capabilities that allow AI agents to interact with operating systems, websites, and applications via mouse, keyboard, and visual input. Developers can use this model to automate multi-step workflows in software environments.
GPT-5.4 supports context windows of up to one million tokens and introduces tool lookup, which allows AI systems to dynamically locate and use external tools without loading all tool definitions into memory. This reduces token usage while improving performance in complex workflows.
Indicators show improvement in a number of tasks. On the GDPval benchmark for professional knowledge work, GPT-5.4 meets or exceeds industry professionals in 83% of cases, compared to about 71% for GPT-5.2. The model also outperforms previous versions in encryption tests, web browsing tasks, and computer usage benchmarks.
The model is also designed to be more efficient. OpenAI says that GPT-5.4 requires fewer tokens to solve many reasoning tasks compared to GPT-5.2, improving response speed and reducing developer costs.
GPT-5.4 Thinking is now available for ChatGPT Plus, Team and Pro users, while GPT-5.4 Pro is available for Pro and Enterprise plans. Developers can access both models via API.
