OpenAI unveils GPT-5: Enhanced AI with PhD-level problem-solving

OpenAI has officially unveiled GPT-5, its latest flagship multi-modal AI model, promising significant advancements in both problem-solving capabilities and code generation. The announcement, made during a livestream, marks a notable shift in accessibility: unlike previous iterations often reserved for paid subscribers, GPT-5 will also be available to free-tier users of ChatGPT.

OpenAI CEO Sam Altman underscored the dramatic leap in capability, likening the experience of interacting with GPT-5 to conversing with a PhD-level expert. He drew a vivid comparison to earlier models, noting that while GPT-3 felt like a high school student and GPT-4 a college student, GPT-5 provides “an entire team of PhD experts in your pocket, ready to help you.” Researchers highlighted that the new model is engineered for enhanced reliability and accuracy, aiming to significantly reduce instances of AI “hallucinations”—erroneous or nonsensical outputs.

Beyond improved reasoning and higher-quality code generation, GPT-5 offers greater autonomy, requiring less user input for complex tasks, and boasts seamless integration with popular platforms such as ChatGPT, Google’s Gmail, and Calendar applications. A key technical evolution in GPT-5 is its ability to incorporate new data from user prompts in real-time, a method known as “test-time computing.” This distinguishes it from older language models that relied solely on pretraining for response generation. This release also unifies OpenAI’s naming convention, consolidating previous model iterations like ‘o1’ and ‘o4-mini’ under the GPT-5 family, signaling a strategic focus on models that seamlessly blend both pretraining and real-time inference.

For many in the AI community, the critical question remains whether the leap from GPT-4 to GPT-5 will prove as transformative as the widely acclaimed jump from GPT-3 to GPT-4. Independent testing will be crucial in evaluating this anticipated performance differential.

In a strategic counterpoint to its typically closed model approach, OpenAI recently introduced two “open-weight” models—gpt-oss-120b and gpt-oss-20b—making them freely available and modifiable by developers. However, with the launch of GPT-5, the company appears to be returning to its more traditional “closed” strategy, maintaining proprietary control over its most advanced technology.

Adding to the day’s significant announcements, OpenAI revealed a landmark partnership with the U.S. federal government. Through an agreement with the General Services Administration (GSA), ChatGPT Enterprise, the company’s advanced chatbot, will be made available to executive branch agencies for an unprecedented rate of just $1 per agency for one year. OpenAI has provided explicit assurances that it will not utilize any government data to train its AI models, addressing critical concerns about data privacy and national security.