OpenAI Launches GPT-5 with Enhanced Capabilities and Safety Features

OpenAI has officially released GPT-5, its latest AI model, now accessible to all users, including those on the free tier, via ChatGPT and OpenAI’s API. The rollout includes specialized variants such as GPT-5-mini, GPT-5-nano, and GPT-5-chat, each tailored for distinct applications. The model promises significant improvements in writing, coding, and healthcare domains, with OpenAI highlighting its reduced hallucination rates—fewer instances of fabricated responses—thanks to 5,000 hours of rigorous safety testing.

Unlike its predecessors, GPT-5 avoids outright refusals to answer sensitive questions. Instead, it employs 'safe completions,' delivering high-level responses within predefined safety boundaries to mitigate potential harm. During a demonstration, OpenAI showcased the model’s 'vibe coding' prowess by tasking it with creating an interactive web app for French language learning. The app, complete with flashcards, quizzes, and progress tracking, was generated in seconds—though OpenAI acknowledged room for refinement.