I wanted to leave a note about how incredibly impressive GPT 4.5 has been for writing (books, letters, emails, EVERYTHING) and for emotional intelligence.
I am excited to share the scope of its efficacy in our API use in experimental products specifically around helping users solve relational conflicts, process difficult emotions, and evaluate or introspect. In addition 4.5 is remarkable at staying consistent with prose, tone, and writing style. This stands true against o3, o4-mini, 4o, and Claude 3.7 at time of writing.
This is a remarkable model. Please don’t deprecate 
A massive amount of our work would be impossible with the existing model set. 4.5 consistently outperforms in writing emotionally intelligent replies with minimal hallucination while rooted in any contextual framework.