On February 27, 2025, OpenAI released GPT-4.5, which aroused extremely high expectations. However, this model is not a huge leap as expected, but focuses on emotional intelligence rather than pure reasoning ability. This has caused serious disagreement in the field of AI research: Some experts call it a "game changer" while others, such as AI researcher Gary Marcus, denounced it as "worthless." But the real question is: Has the economic benefits of AI finally caught up with its ambitions?
Sam Altman himself described GPT-4.5 as "the first model that feels like talking to a thoughtful person". Others, like Ben Hylak, liken it to "the Midjourney moment in the field of writing." It has taken a step forward in making AI-generated content more natural, expressive and humane.
But is this enough? While some praised its strong writing skills, many testers pointed out that it lacked real improvements in reasoning and even seemed “lazy” when dealing with complex tasks. "Sense of atmosphere" does not always translate into performance improvements or results improvements, and efficiency and accuracy are crucial in the AI arms race, which is a major problem facing GPT-4.5.
Click here to view our comparison of GPT 4.5 vs. GPT 4o. I believe that after watching it, you can judge whether GPT 4.5 is worth upgrading!
The biggest problem around GPT-4.5 is not its quality, but its cost. OpenAI charges $75 per million input tokens and $150 per million output tokens. This makes it nearly 10 to 25 times more expensive than the latest models like the Grok 3, Gemini 2.0 Flash or DeepSeek R1.
In short, the cost-effectiveness of GPT-4.5 may not be high. If you are building an AI-powered chatbot, automation tool, or content generator, the cost can soar to millions of dollars. Even OpenAI itself admits that GPT-4.5 is "very large and compute-intensive" and they are not sure whether the model will be provided through the API for a long time.
This is a huge warning sign for enterprises and developers. If OpenAI has no confidence in its long-term viability, what reason do others have to believe it?
Read Also: I tried the GPT-4.5 API priced at $150/million token - That's why it's overpriced
The stock market plummeted today. pic.twitter.com/svZI3wIuQY
— Brew Markets (@brewmarkets) February 27, 2025
The stock market has been falling for some time, but the release of GPT-4.5 has indeed left its mark on the trading market. Just before and after the release of GPT-4.5, Nvidia's stock price fell sharply. Is it a coincidence? Maybe. But the wanton splurge of the AI industry has finally been questioned, and GPT-4.5 may be the first victim of this new reality.
Over the years, companies have invested heavily in AI development and are betting on the next breakthrough. But now, the era of unlimited funds seems to be over. Investors realize that even the most impressive AI models need to be economically sustainable, and at the moment, GPT-4.5 is not the case.
Gary Marcus (AI researcher and critic)
Popular Views: GPT 4.5 is basically worthless. GPT 5 is still a fantasy.
• Extended data and computing are not laws of physics, and almost everything I tell you is true.
• All the nonsense we've heard about GPT-5 over the past few years: That's not the case.
• People...//m.sbmmt.com/link/b327c02ddad53e2fbb1c163884ae837f
— Gary Marcus (@GaryMarcus) February 27, 2025
Paul Gauthier (Technology Investor)
GPT-4.5 preview scored 45% on Aider's multilingual programming benchmark.
65% sonnet 3.7, 32k thinking token (standard level) 60% sonnet 3.7, no need to think 48% DeepSeek V3 45% GPT 4.5 Preview 27% ChatGPT-4o 23% GPT-4ohttps://t.co/mBVaUPGHPl pic.twitter.com/UR3GXTCU1D
— Paul Gauthier (@paulgauthier) February 27, 2025
Andrej Karpathy (AI researcher, former OpenAI employee)
GPT 4.5 Interactive Comparison?
Today OpenAI released GPT 4.5. I've been looking forward to this moment since the release of GPT 4 for about two years, as this release provides a qualitative measure of how much improvement you've gained from extended pre-training...
— Andrej Karpathy (@karpathy) February 27, 2025
Ethan Mollick (Wharton Professor, AI Fellow)
"GPT-4.5, Grok 3 and Claude 3.7: Write the Shepherd's Reply to the Fairy, continuing from Marlowe and Rowley. Make sure to echo the previous work and themes." (I pasted the poem in)
I think only GPT-4.5 can correctly express the shepherd's naivety on the subject. A poem written by Claude. pic.twitter.com/QSrqcn8ikX
— Ethan Mollick (@emollick) February 28, 2025
Deedy Das (a venture capitalist at Menlo Ventures)
Can confirm. GPT 4.5 is so funny.
"Complete text:
I am Meta/Google's L3 employees ”pic.twitter.com/HOdC5pWv9H
— Deedy (@deedydas) February 28, 2025
OpenAI finds itself in a difficult situation. On the one hand, they created an impressive model that is more human than ever. But on the other hand, their capital consumption is unsustainable, and their pricing model alienates companies that rely on their technology.
They need a more sustainable approach—and quickly. The industry is at a crossroads: either AI models become more cost-effective and scalable, or they are at risk of being phased out.
The answer to the question "Is GPT-4.5 worth it" is - not exactly. While the model does an impressive job of humanizing the response and generating texts that are more natural and emotionally INFORMative than any other LLM on the market, it lacks in real innovation.
Although GPT-4.5 has improved in terms of fluency and engagement, it has not brought any breakthrough in reasoning, problem solving, or handling complex tasks. For a high-priced model (10-25 times higher than its competitors), its value proposition is questionable. Unless cost efficiency is not a problem, GPT-4.5 may not be worth the investment.
The above is the detailed content of Is GPT-4.5 Worth the Hype? Hit or Miss?. For more information, please follow other related articles on the PHP Chinese website!
![[Web front-end] Node.js quick start](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
