OpenAI Launches GPT-5.4 AI Models with powerful upgrades. This release improves reasoning, coding, and agent workflows. The company announced the new model on Thursday. It is the third major update of GPT-5.
The model works across ChatGPT, API, and Codex platforms. OpenAI also released a stronger GPT-5.4 Pro model. The update introduces computer-use capabilities for AI agents. These features allow AI to interact with software.
also read: How OpenAI Integrated ChatGPT with WhatsApp: Full Guide 2026
OpenAI Launches GPT-5.4 AI Models Across Platforms
The model is available inside ChatGPT and developer tools. Developers can also access the model using API. The model is integrated with the Codex platform.
OpenAI named the main version GPT-5.4 Thinking. Another advanced version is called GPT-5.4 Pro.
At launch some users receive access first. Plus, Team, and Pro subscribers get GPT-5.4 Thinking. This version replaces the older GPT-5.2 Thinking model.
The stronger GPT-5.4 Pro targets advanced professional usage. Pro and Enterprise subscribers can use this version.
The new model combines many OpenAI improvements together. It includes coding ability from the GPT-5.3 Codex system. The model works better with tools and workflows.
It can handle spreadsheets, documents, and presentations easily. The system also supports extremely large context windows. GPT-5.4 can process up to one million tokens.
This helps analyze huge codebases or document collections. Developers can review long files without splitting data.
With Computer Use
This feature allows AI agents to control computers. The system interacts with apps and websites directly. It uses screenshots to understand the computer interface.
The model can send mouse clicks and keyboard commands. AI agents can browse websites and complete actions. They can even place online orders automatically. The system writes code that controls computer operations.
OpenAI tested this ability using OSWorld benchmark tests. The model scored seventy-five percent success in testing. This benchmark measures navigation in desktop environments. The AI reads screenshots and performs correct actions.
The upgrade also improves visual understanding capabilities. Visual perception helps AI understand computer interfaces better. The model scored strongly on the MMMU-Pro benchmark. It achieved eighty-one percent success in visual reasoning.
The system supports long tasks and multi-step workflows. AI agents can plan tasks and execute them. They also adjust actions if problems appear.
For Professional Work
The system performs strongly across complex business tasks. It can analyze financial spreadsheets and company data. Businesses can build dashboards automatically using the model.
The model also drafts reports using raw data. It can process huge legal contracts and documents. Law and finance teams may benefit from this capability.
Developers receive powerful improvements for coding workflows. The model writes large software codebases efficiently. It can debug programs and detect coding errors. The system also runs automated testing processes.
Another upgrade reduces false answers from AI systems. OpenAI claims thirty-three percent fewer hallucination errors. This improvement increases reliability for professional tasks.
The company also added strong safety protections. GPT-5.4 is classified under high cyber capability risk. OpenAI deployed an expanded cyber safety monitoring system. Safety tools include monitoring, trusted access, and blocking.
OpenAI also introduced chain-of-thought controllability evaluation. This test checks if models hide reasoning intentionally.
Meanwhile competition in AI technology continues growing. Anthropic recently released the Claude Opus 4.6 model. The company also launched the Claude Sonnet l 4.6 model. These models target enterprise workloads and productivity tasks.
OpenAI Launches GPT-5.4 AI Models with major improvements. Earlier AI mostly answered questions and generated text. Now models can perform tasks on computers directly.
Businesses may use AI agents for daily workflows. The release also shows growing competition in the AI industry. Future AI systems may become powerful digital workers.
