OpenAI has officially launched GPT-4.5, its largest and most knowledgeable model to date, available as a research preview. While boasting improved writing capabilities and refined personality traits over its predecessors, OpenAI cautions that GPT-4.5 is not a frontier model and performs below certain models like o1 and o3-mini. Improved computational efficiency exceeding 10x has been noted, and the model is reported to hallucinate less than earlier versions. Access is expected to be limited to ChatGPT Pro users, concluding with hints at the upcoming GPT-5 release.
GPT-4.5 is OpenAI's largest LLM, improving on GPT-4's computational efficiency by more than 10x, but it's not a frontier model.
While GPT-4.5 improves writing capabilities and knowledge, it doesn't introduce new frontier capabilities and performs below o1 and o3-mini.
Despite its limitations, GPT-4.5 hallucinates much less than GPT-4o and slightly less than the o1 model, showcasing improved reliability.
OpenAI plans to limit GPT-4.5 access to ChatGPT Pro users, similar to the release strategy for Operator and has future plans for GPT-5.
Collection
[
|
...
]