OpenAI introduces ChatGPT-4.5, an AI model with advanced emotional intelligence, but CEO Sam Altman warns that its training and operation come at a significant cost.
 |
| OpenAI's new GPT-4.5 is more emotionally intelligent and adept at social cues, but Sam Altman cautions that the model’s training and operation remain costly. Image/ Illustration: ChicHue |
San Francisco, USA — February 28, 2025:
OpenAI has officially introduced GPT-4.5, its most advanced AI model to date, which promises to revolutionize the way AI interacts with humans. Described by CEO Sam Altman as “the first model that feels like talking to a thoughtful person,” GPT-4.5 is designed to be more emotionally intelligent and context-aware than its predecessors.
In a post on X (formerly Twitter), Altman expressed his amazement at the model’s capabilities, saying, “I have had several moments where I’ve sat back in my chair and been astonished at getting actually good advice from an AI.” The new model is being heralded as a major leap in AI development, offering users more natural and emotionally nuanced conversations.
However, while GPT-4.5 represents a significant advancement, Altman was quick to warn that the new model is “giant” and “expensive.” The cost of training and running such a powerful model has been a key concern for OpenAI. Although the company has reduced the operational costs of large-scale models, Altman emphasized that the expense of GPT-4.5’s training remains high. Due to limitations in hardware, OpenAI’s rollout of the model to its wider user base was delayed, with Altman stating that the company ran out of GPUs. “We will add tens of thousands of GPUs next week and roll it out to the plus tier then,” he said.
OpenAI's focus on scaling models through increased data and computational power seems to be paying off, as the company claims to have not yet seen diminishing returns from scaling. "We’re very rigorous about how we do this," said OpenAI's Chief Research Officer Mark Chen. "This is the point that lies at that next order of magnitude."
GPT-4.5 is designed to excel in emotionally intelligent tasks, making it more adept at responding to social cues than earlier models like GPT-4o. In a live demonstration, OpenAI staff showed how GPT-4.5 could reframe a confrontational text to a friend, using tact and empathy. This emotional nuance is one of the model’s core strengths, making it a more effective and intuitive tool for a wide range of applications, including writing, programming, and problem-solving.
Additionally, GPT-4.5 builds on OpenAI’s “unsupervised learning” paradigm, which focuses on scaling models to enhance word knowledge, intuition, and reduce hallucinations. According to OpenAI, the model’s stronger alignment with user intent and its vast knowledge base enable it to handle tasks more accurately and creatively than previous versions.
GPT-4.5 has also undergone rigorous testing against a "vibes" test set that measures emotional and creative intelligence. It outperformed OpenAI's previous multimodal models, like GPT-4o, showing its ability to tackle not just technical problems but also those that require a softer touch.
Although OpenAI has made great strides in improving the model’s abilities, the cost of scaling up these advanced models remains a challenge. In response, OpenAI has worked on optimizing how it runs large models, reducing inference costs significantly since the release of GPT-4.
For now, GPT-4.5 is available to users who subscribe to ChatGPT Pro at $200 per month, with plans to expand to other subscription tiers, such as ChatGPT Plus, Team, and Edu, in the coming weeks. Despite the high operational costs, OpenAI’s push for more advanced AI models signals a future where the line between human and machine interactions continues to blur, offering new opportunities while raising questions about the sustainability of such rapidly evolving technology.
As the debate over the cost and impact of AI continues, OpenAI’s latest release underscores both the potential and the challenges of developing emotionally intelligent, powerful artificial intelligence. With the global community watching closely, it remains to be seen how OpenAI will balance innovation with the costs required to maintain such advanced systems.