OpenAI’s GPT-4.5 might not outperform in benchmarks, but it aims to enhance user connection
OpenAI’s latest preview of GPT-4.5 reportedly improves on past models by enhancing emotional intelligence and advancing creative capabilities.
Critics have voiced concerns that the model costs too much.
While OpenAI’s next model could be significantly more expensive—potentially 20 times the cost of its predecessor—early feedback suggests it will foster far more creativity and natural engagement in conversations.
OpenAI introduced GPT-4.5 as a research preview on Feb. 27, describing it as the company’s most advanced AI. The model recognizes patterns, forms connections, and generates creative insights, though it does not significantly improve reasoning compared to previous versions.
According to OpenAI’s Feb. 27 announcement, GPT-4.5’s greater knowledge and heightened emotional intelligence allow it to handle creativity and practical problem-solving more effectively.
“We also expect it to hallucinate less and deliver more reliable performance across a wide range of general topics, including richer conversations.”
While GPT-4.5 excels in creativity and conversational flow, OpenAI stated that it does not process intricate logical tasks as effectively as the o-series models.
OpenAI noted that GPT-4.5 has advantages but struggles with step-by-step reasoning and slower processing due to its complexity. The model does not generate audio or video content.
Dan Shipper, CEO of AI and publisher of the business newsletter ‘Every’, reviewed OpenAI’s newest model and shared observations similar to others.
Shipper, who has been testing the updated version with his team, commented, “It’s not going to blow your mind, but it might befriend you.”
“It’s more like a personality, communication, and creativity upgrade than a huge intelligence leap. It’s like OpenAI is pivoting its base model from ‘bland assistant’ to ‘AI bestie.’”
Shipper observed that GPT-4.5 sometimes struggles to follow given instructions.
AI expert Aran Komatsuzaki reported that GPT-4.5’s API usage costs significantly more, approximately 15 to 20 times the price of GPT-4o.
In a Feb. 27 X post, OpenAI’s Sam Altman acknowledged that the new reasoning model operates as a “giant expensive model” and “won’t crush benchmarks.”
Some experts, like biomedical scientist Derya Unutmaz of The Jackson Laboratory, praised GPT-4.5’s medical imaging abilities, highlighting its accurate diagnosis of a tubal ectopic pregnancy.
Professor Unutmaz pointed out that GPT-4.5 correctly diagnosed the condition, whereas models like Grok 3, Claude 3.7 Sonnet, Gemini 2.0, and older ChatGPT versions misinterpreted the medical image as a normal pregnancy.
OpenAI’s newest version of ChatGPT enters the market just as High Flyer, a China-based company, launched DeepSeek R1 in January—an open-source AI model developed at a far lower expense than OpenAI’s.
OpenAI’s chief executive, Sam Altman, argues that AI model production costs are rapidly declining, reducing by at least ten times annually.
“You can see this in the token cost from GPT-4 in early 2023 to GPT-4o in mid-2024, where the price per token dropped about 150x in that time period,” Altman said in a Feb. 10 post.
OpenAI revealed on Feb. 13 that GPT-5 will unify several versions, including o3. On Feb. 12, Altman stated that GPT-5 will arrive within months.
Users on the free version of ChatGPT will be able to chat with GPT-5 without restrictions.
Four months ago, OpenAI launched the ChatGPT Search to simplify real-time access to information.
The product empowers users to swiftly scour the web and fetch quick answers, complete with links to pertinent sources.
Users can now get real-time updates on sports scores, news, and stock prices straight in their chat, bypassing the need for standard search engines.
The feature can either automatically search based on a user’s query or lets users manually trigger a search by clicking an icon. It’s accessible via chatgpt.com and available on both desktop and mobile apps.
