OpenAI has introduced GPT-5, the latest version of the artificial intelligence system behind ChatGPT, arriving more than two years after GPT-4’s release in March 2023. The announcement comes during a period of intense competition, with companies like Anthropic, Google, and several Chinese AI developers releasing upgraded models in quick succession.
Chief Executive Sam Altman described GPT-5 as a major advancement in capability and reliability. He compared the change from earlier models to moving from conversations with a high-school student to discussions with an expert holding a doctoral degree. OpenAI reports that the system delivers faster, more accurate answers and has improved reasoning skills, particularly in software development, professional writing, and medical-related queries.
One of the most visible changes for users is how the model operates within ChatGPT. Instead of requiring people to select between a standard model and a reasoning model, GPT-5 uses an automatic “router” that detects complex prompts and switches to the reasoning mode without manual intervention. The system is available now to all ChatGPT accounts, though free users face a daily usage limit before the service defaults to a lighter version. Developers can connect to GPT-5 through the OpenAI API, with three distinct options: GPT-5, GPT-5 mini, and GPT-5 nano, each aimed at different cost and performance needs.
The update adds customization tools, including four personality modes—“Cynic,” “Robot,” “Listener,” and “Nerd”—and the ability to assign colors to individual chat threads. OpenAI says the model outperforms its predecessors in widely used programming benchmarks such as SWE-Bench and SWE-Lancer. In one demonstration, it generated a complete interactive language-learning website, producing hundreds of lines of working code within seconds.
Safety improvements were a key focus during development. According to OpenAI, GPT-5 underwent more than 5,000 hours of testing to identify and reduce risks. The company has worked to make the model less likely to produce misleading or harmful outputs, especially in cases where prompts are intentionally phrased to bypass safeguards. A new “safe completion” feature allows the model to offer helpful, higher-level information for potentially dangerous topics while withholding technical details that could cause harm.
Microsoft is deploying GPT-5 across a range of products, including Microsoft 365 Copilot, Microsoft Copilot, GitHub Copilot, Visual Studio Code, and Azure AI Foundry. This integration will allow users to access the new reasoning capabilities without needing to choose the best model themselves. Microsoft’s internal AI Red Team, which stress-tests systems for potential misuse, reported improved resistance to threats such as malicious code generation and scam automation.
With GPT-5 in Microsoft 365 Copilot, users can work through complex problems, sustain focus in extended conversations, and analyze workplace documents in greater depth. Developers using GitHub Copilot will benefit from faster, more precise code generation, while Azure AI Foundry clients can take advantage of the router’s ability to select the most efficient model for each task.
Although OpenAI views GPT-5 as an important advancement toward artificial general intelligence, Altman has acknowledged it still lacks continuous learning—something he believes is essential for future systems. In a fast-moving AI market, the company’s challenge will be to keep advancing technology while adapting to increasing competitive and pricing pressures.
This image is the property of The New Dispatch LLC and is not licenseable for external use without explicit written permission.
