OpenAI has released its largest and most powerful model to date, GPT-4.5, which the company claims feels like talking to a thoughtful person. CEO Sam Altman described it as “giant” and “expensive,” but assured that the company will add more GPUs next week and roll out the model to paid subscribers.
The new model has been trained on vast amounts of data and computational resources, which have led to significant improvements in its performance. According to Mark Chen, OpenAI’s chief research officer, the company has not seen diminishing returns from scaling up its models, and GPT-4.5 marks a major breakthrough.
GPT-4.5 is designed to think before responding, making it better suited for quantitative tasks. However, it excels in conversational and emotionally intelligent interactions, picking up on social cues and responding more aptly than other models. In tests, GPT-4.5 outperformed its predecessor, 4o, with fewer hallucinations and improved accuracy.
The model’s enhanced knowledge base and alignment with user intent make it well-suited for tasks like writing, programming, and solving practical problems. While training costs remain high, OpenAI has found ways to reduce inference costs, making the model more accessible to a wider range of users.
GPT-4.5 is currently available in a research preview to ChatGPT Pro and API developers, with plans to roll out the model to other subscribers next week.
Source: https://www.businessinsider.com/openai-sam-altman-releases-gpt-4-5-emotionally-intelligent-model-2025-2