OpenAI launches GPT-5, a "PhD-level expert" with a unified approach to AI

OpenAI has officially launched GPT-5, a new flagship model that the company is touting as its most capable and intelligent system yet. The release, which comes more than two years after the debut of GPT-4, is being closely watched as a bellwether for the future of generative AI.

According to OpenAI, GPT-5 represents a significant leap forward, particularly in its reasoning, coding, and writing abilities. OpenAI CEO Sam Altman described the new model as akin to a "PhD-level expert," a substantial improvement over GPT-4's "college student" feel.

The company also claims to have made strides in reducing "hallucinations," with a 45% reduction in factual errors compared to its predecessor.

One of the most notable changes is GPT-5’s new unified system. The model can now intelligently route user requests, automatically deciding whether to provide a fast response or engage in deeper, more complex reasoning. This eliminates the need for users to manually select different models for different tasks, offering a more streamlined experience.

Following the launch, Elon Musk, CEO of xAI, quickly took to social media to claim that his company's Grok 4 Heavy AI is still the most powerful while announcing the arrival of Grok 5 by year-end. This public sparring highlights the intense competition at the forefront of the AI race.

User reactions and new features

Initial user reactions have been largely positive, with many developers and early testers praising the model's improved capabilities. For coders, GPT-5's "vibe coding" feature, which allows for the creation of functional applications from simple prompts, is a standout.

The model also demonstrates significant improvements in front-end development, resulting in cleaner and more aesthetically pleasing interfaces.

For everyday users, the new model offers several key updates:

• Multimodality: GPT-5 is a unified system for text, images, voice, and live video.

• Customization: Users can now set a "personality" for their chatbot, with options like "sarcastic" or "professional," and even change the chat bubble colors.

• Integrations: The model can connect to services like Gmail and Google Calendar, enabling it to manage schedules and draft emails.

• Enhanced Safety: OpenAI has implemented new safety measures to make the chatbot "less deceptive" and to provide safer responses to potentially dangerous queries.

While some experts remain skeptical about the pace of AI development and the ethical implications, GPT-5's launch marks a new chapter in the ongoing quest to create a more intelligent and useful artificial intelligence.