News Summary

OpenAI’s latest AI model, GPT-4.5, has been unveiled to mixed reviews. The model, which is touted as the company’s most advanced traditional AI model to date, has been criticized for being expensive and slow, offering only marginally better performance than its predecessor, GPT-4o, while costing significantly more. Despite the high expectations, GPT-4.5 is noted for its subtle improvements over previous models, as highlighted by former OpenAI researcher Andrej Karpathy. However, benchmark tests show it scores lower on standardized assessments compared to other OpenAI models, raising questions about its cost-effectiveness and practical utility. OpenAI has acknowledged the model’s limitations and is evaluating its long-term viability in the API. The release has sparked discussions about the diminishing returns in training large unsupervised learning models.