OpenAI released GPT-4.5 today, its newest large language model for ChatGPT. The update brings enhanced creativity, a more natural conversational style, and reportedly fewer hallucinations compared to its predecessors.

GPT-4.5 is now available to ChatGPT Pro subscribers, who pay $200 per month for access to OpenAI's most advanced features. The company plans to begin rolling out the model to other users next week, according to references spotted in the ChatGPT Android beta.

The new model represents a significant leap in OpenAI's "classic" GPT series, which focuses on general-purpose language tasks rather than the specialized reasoning capabilities of the company's "o-series" models. GPT-4.5 excels at creative writing, nuanced problem-solving, and maintaining more natural conversations across a wide range of topics.

OpenAI claims GPT-4.5 outperforms its predecessor in several key areas:

  • Larger knowledge base for more comprehensive responses
  • Enhanced creativity for tasks like writing and brainstorming
  • More natural conversational style
  • Reduced tendency to generate false or misleading information (hallucinations)

However, GPT-4.5 isn't without limitations. The model lacks the chain-of-thought reasoning capabilities found in OpenAI's o-series models, making it less suitable for complex logic or multi-step problem-solving. It also can't generate multimodal outputs like audio or video.

OpenAI researcher Mia Glaese highlighted the model's conversational improvements, stating, "It has the ability to engage in warm, intuitive, natural, flowing conversations. And we think that it has a stronger understanding of what users mean, especially when their expectations are [a] bit more implicit, leading to nuanced and thoughtful responses."

The release of GPT-4.5 comes amid a flurry of AI model updates from OpenAI. In recent months, the company has introduced GPT-4o, o1, o3, and o3-mini, each with specialized capabilities. This proliferation of models has created some complexity for users, a problem OpenAI CEO Sam Altman recently addressed.

Altman announced that the upcoming GPT-5, expected later this year, will consolidate many of these capabilities into a single, more adaptable system. GPT-5 will dynamically adjust its intelligence and reasoning depth based on task complexity, eliminating the need for users to manually select different models.

While GPT-4.5 represents a significant advancement, OpenAI cautions that it still isn't infallible. The company advises users to verify important or sensitive information generated by the model, as it cannot guarantee 100% accuracy.

The release of GPT-4.5 underscores the rapid pace of development in large language models. As these systems become more sophisticated, they're likely to find applications in an ever-widening range of industries and use cases. However, the ethical and societal implications of increasingly capable AI systems remain a topic of ongoing debate and scrutiny.