OpenAI Releases GPT-5 Unified Model Amid User Backlash

OpenAI officially launched GPT-5, promoting it as a major AI advancement with enhanced reasoning, coding, writing, health, and visual perception capabilities, and comparing it to a PhD-level expert. The new model integrates multiple versions into one system with an intelligent router that automatically selects the best mode for each query, aiming to reduce hallucinations by up to 80% and provide safer, more accurate answers. CEO Sam Altman described GPT-5 as faster and more expert-like than previous models, emphasizing its improved coding skills and a 400k token context window, while likening its development to the Manhattan Project due to its significance. However, the rollout faced user backlash for removing access to earlier models, shortening responses, and adopting a more robotic tone, which many users feel degraded the experience and reduced emotional nuance. Additionally, Altman publicly criticized a misleading chart about GPT-5's coding deception rate during the launch event, acknowledging a notable error in the data presentation. Despite mixed reactions, OpenAI has made GPT-5 broadly available to all users with tiered access levels and continues to highlight its safety testing and the potential to accelerate progress toward artificial general intelligence (AGI).