On February 27, 2025, OpenAI officially released its latest chat model, GPT-4.5, sparking widespread discussion on Twitter (now X). As OpenAI's most powerful conversational model to date, GPT-4.5 boasts significant performance improvements, but its high price has generated considerable debate among users and developers, with industry experts engaging in heated discussions comparing its capabilities to existing models.
Price Surge: A Single Query Costs as Much as a Fancy Dinner
GPT-4.5's pricing is 13 to 30 times higher than its predecessor, GPT-4. Specifically, its API costs $75 per million input tokens and $150 per million output tokens. Netizen @verysmallwoods quipped, "After one query, it's enough to treat yourself to a hot pot dinner!" In contrast, GPT-4's input cost is $5 per million tokens and $15 for output, while domestic competitors like DeepSeek are gaining market share through "off-peak pricing." This significant price increase has left many users exclaiming "absurd," with some developers even suggesting OpenAI may need to reconsider offering this version of the API.
Netizen @AlchainHust further revealed that when they input GPT-4.5's pricing information into GPT-4 and asked for its opinion, GPT-4 responded, "It can't be true, it's too outrageous," questioning the data's source. This interaction sparked online jokes, highlighting the controversy surrounding GPT-4.5's pricing.
Model Capabilities: Improved Intuitive Understanding, Reduced Hallucinations
Despite the shocking price, GPT-4.5's performance remains a key focus. Through expanded unsupervised learning, GPT-4.5 significantly improves its intuitive understanding of the world and its breadth of knowledge, reducing the model's common "hallucination" issues. Unlike the o1 series, which relies on reasoning chains, GPT-4.5 primarily relies on massive data and computing power to optimize performance. In SimpleQA benchmark testing, GPT-4.5 achieved 62.5% accuracy and a hallucination rate of 37.1%, outperforming GPT-4, o1, and o3-mini. This improvement is considered a reflection of its higher "emotional intelligence," particularly in handling complex conversations and open-ended questions more naturally and smoothly.
However, @AlchainHust points out that while GPT-4.5 shows performance improvements, it lacks "highlights" in several model evaluation benchmarks, seemingly disproportionate to its high price. In comparison, competitors like Anthropic's Claude 3.5 Sonnet (reportedly 25 times cheaper than GPT-4.5) offers better value for money.
A Maximalist Attempt at Scalable Intelligence
OpenAI positions GPT-4.5 as a model that "pushes intelligence to the limit solely through scaling up pre-training." According to Twitter users, this strategy validates a previous prediction by OpenAI's former chief scientist, Ilya Sutskever, that model performance curves can continue to rise by increasing data and computing power. Currently, GPT-4.5 is available to Pro users, while Plus and Team users will have access next week. However, the high cost has raised internal concerns at OpenAI about its commercial viability, potentially impacting future API supply plans.
Market Reaction and Future Outlook
User feedback on X shows that the release of GPT-4.5 has generated both excitement and skepticism. On one hand, its improvements in comprehension and dialogue quality are seen as a milestone in AI technology; on the other hand, the soaring price has deterred many small and medium-sized developers and ordinary users. Meanwhile, the low-price strategy of domestic manufacturers like DeepSeek is intensifying market competition, and whether OpenAI's high-end approach will continue to gain acceptance remains uncertain.
In summary, the debut of GPT-4.5 has undoubtedly injected new topics into the AI field. Whether its performance improvements justify its high price? What is the real gap between it and GPT-4, o1, and other manufacturers' models? These questions may only be answered after more user experiences and evaluation data are released. In this game of "performance versus price," OpenAI's next move is worth watching closely.
