OpenAI has unveiled GPT-4.5, its latest and most advanced AI model, currently available to select paying users. This release is part of a research preview, initially accessible to ChatGPT Pro subscribers at $200 per month and developers on paid API tiers.
GPT-4.5 introduces several enhancements, including improved pattern recognition, the ability to generate creative insights without explicit reasoning, and heightened emotional intelligence. These advancements aim to provide more natural and thoughtful interactions, making the AI feel more like engaging with a considerate individual.
One notable improvement is the significant reduction in the model’s “hallucination” rate—the occurrence of generating incorrect or nonsensical information. GPT-4.5 has reduced this rate to 37.1%, down from 61.8% observed in previous models, enhancing the accuracy and reliability of its responses.
However, the model’s substantial size and computational demands have led to challenges in its broader deployment. OpenAI CEO Sam Altman acknowledged that GPU shortages have limited the immediate availability of GPT-4.5, restricting its release to higher-tier subscribers initially. The company plans to acquire additional GPUs to expand access to ChatGPT Plus and Team users next week, followed by Enterprise and Edu users subsequently.
Despite these advancements, GPT-4.5 may not achieve record-breaking benchmarks. Altman emphasized that while the model represents a significant step forward, it might be the last of its kind before a strategic shift towards improving AI reasoning capabilities, moving away from the “bigger is better” approach.
In summary, GPT-4.5 marks a pivotal development in OpenAI’s pursuit of more human-like and emotionally intelligent AI interactions, with plans to broaden access as infrastructure capabilities are enhanced.
