OpenAI has launched GPT-5.4, its newest frontier artificial intelligence model, introducing major upgrades in reasoning, coding and automated task execution.
The company said the model combines several of its recent advancements into a single system and is available in different variants, including GPT-5.4 Thinking and GPT-5.4 Pro.
One of the most significant features of GPT-5.4 is its 1 million-token context window, allowing it to analyse very large datasets such as entire codebases or extensive collections of documents more efficiently.
OpenAI also said GPT-5.4 is the first mainline model with built-in computer-use capabilities, enabling AI agents to directly interact with software to complete tasks. This means the system can operate computers by using screenshots, mouse clicks and keyboard commands, allowing it to work across applications and websites and automate complex workflows.
According to the company, the latest model introduces six major improvements, including enhanced coding abilities, better image perception and multimodal performance, stronger execution of long-running tasks and multi-step agent workflows, improved token efficiency for tool-heavy workloads, advanced web search and multi-source information synthesis, and more effective document-heavy analytics.
Addressing concerns about inaccuracies often referred to as “hallucinations,” OpenAI said GPT-5.4 is 33% less likely to produce false information compared with earlier models.
The company said the model is designed for professional environments and performs strongly in tasks such as legal analysis, financial modelling, creating presentation slides and writing or debugging code. Developers can also build AI agents capable of planning tasks, carrying them out and adjusting when problems arise.
The release reflects a broader shift in the evolution of AI systems. Early versions of ChatGPT primarily answered questions, while the GPT-4 era enabled more advanced capabilities such as writing essays, code and summaries. With GPT-5, models began to demonstrate stronger reasoning skills, and GPT-5.4 moves further by allowing AI systems to directly perform tasks on computers.
In practical use, GPT-5.4 can operate within common workplace tools such as spreadsheets and document editors. It can analyse financial data in Excel, automatically create dashboards, generate reports from raw datasets and process large legal or contractual documents.
For software development, the model can generate extensive codebases, detect and fix bugs, run automated software tests and even control web browsers through automation tools.
OpenAI’s latest release comes amid intensifying competition in the AI sector. Rival company Anthropic, led by Dario Amodei, recently introduced Claude Opus 4.6 and Claude Sonnet 4.6, which have been described as faster and more efficient for everyday enterprise tasks.
While the latest models from OpenAI and Anthropic focus on different strengths, the developments highlight a growing race to create AI systems capable of functioning as practical digital workers.
#From Indian Express