OpenAI's GPT-4.5 Faces Criticism Amid Shift to Hybrid AI Models

OpenAI's new AI model, GPT-4.5, has received mixed feedback, particularly regarding its underwhelming performance compared to earlier models. CEO Sam Altman announced that GPT-4.5 represents the final traditional model, with future development shifting towards a hybrid of large language models and simulated reasoning models like o3. Concerns have also arisen over GPT-4.5's pricing, which is significantly higher than GPT-4o, making it less viable for many developers, especially when alternatives like the more affordable o1 pro and o3-mini models offer better performance for certain tasks. This change indicates OpenAI's acknowledgment of the limitations of scaling up LLMs, prioritizing innovative inference methods instead.

Opens on Ars Technica