Recently, tech media outlet The Decoder published an article about OpenAI's latest model, GPT-4.5, sparking industry debate about its cost-effectiveness. Official data shows that while GPT-4.5 offers performance improvements over its predecessor, GPT-4, the cost has increased significantly.
Specifically, GPT-4.5 outperforms GPT-4 in several areas: professional queries (63.2%), everyday queries (57%), and creative tasks (56.8%). However, these improvements range only from 6.8% to 13.2%. More striking is the cost: GPT-4.5 charges $75 per million input tokens and $150 per million output tokens. In contrast, GPT-4 costs $2.5 for input and $10 for output. This means GPT-4.5 is 30 times more expensive for input and 15 times more expensive for output.
OpenAI states that GPT-4.5 won't replace GPT-4 and is evaluating the long-term feasibility of offering it via API. They are also gathering user feedback to determine if GPT-4.5 offers enough unique value to justify its high operational costs. It's worth noting that if GPT-4.5 is ultimately discontinued, application developers relying on it may face challenges.
To address these challenges, OpenAI might release an optimized version, perhaps GPT-4.5 Turbo, to reduce costs. However, given the limited performance improvements of GPT-4.5, whether this will offer a significant advantage remains questionable. Past experiences with GPT-4 Turbo, while faster and cheaper, generally showed lower quality than the original GPT-4.
Amidst the rapid advancement of AI technology, OpenAI faces the significant challenge of balancing model performance and cost. Finding a viable path forward will directly impact its market competitiveness and user adoption.
