Sam Altman announced the release of GPT-4.5 and the upcoming GPT-5 on February 12, generating excitement due to GPT-5 being free and replacing previous models. Meanwhile, the anticipated o3 model won’t be launched, possibly due to energy consumption concerns and competition from DeepSeek-R1. OpenAI introduced o3-mini for cost-effective functionality, indicating that future models will integrate advanced capabilities initially intended for the o series, rather than introducing a standalone o4.
Exciting News: The Arrival of GPT-4.5 and GPT-5
In a surprising turn of events, Sam Altman took to Twitter on February 12 to unveil the highly anticipated GPT-4.5 (codenamed Orion) and the forthcoming GPT-5 models of ChatGPT. This announcement has created quite a buzz, as Altman stated that GPT-5 would be free to use and would replace all prior models, streamlining functionality by automatically activating the appropriate “agents” such as Search and Tasks.
Shifting Focus: The O3 Model Dilemma
By revealing GPT-5 after a year of speculation, Altman may have successfully shifted attention away from another significant announcement: the o3 model, touted as the most powerful AI framework to date, will not be released as initially planned. This raises questions about whether Altman strategically used the launch of GPT-4.5 and GPT-5 to divert focus from the cancellation of o3.
Concerns regarding the energy consumption of the o3 model may have influenced OpenAI’s decision, especially in light of the emerging competition from DeepSeek-R1. In a detailed tweet, Altman addressed the need for OpenAI to enhance transparency and simplify its offerings, noting that the upcoming GPT-4.5 would be followed by GPT-5, which aims to unify traditional language models and advanced thinking models.
Altman’s statement that “we will not release o3 as an independent model” is particularly telling. It suggests that GPT-5 will inherently integrate o3’s capabilities when required, rather than making it a standalone feature. Simultaneously, the announcement of GPT-5 being free may have further distracted audiences from the implications of o3’s discontinuation.
The abandonment of the o3 model, which was expected to be the final breakthrough before achieving general AI, likely stems from financial considerations. Early testers have reported that o3 is exceptionally intelligent but also comes with a hefty price tag—potentially over 1,000 euros per query at full capacity. Consequently, launching such an expensive model would be economically unfeasible.
Moreover, the competition posed by DeepSeek-R1, which offers a significantly lower cost per query, adds to the challenges OpenAI faces. Although o3 would likely outperform DeepSeek-R1, the cost disparity makes it impractical for market release. Additionally, the French company Mistral is strategically focusing on low operating costs for its future models.
Considering the powerful capabilities of o3, did OpenAI intentionally raise expectations among tech enthusiasts? While it remains the leading AI model globally, its optimization still leaves room for improvement. Altman has hinted at various versions of GPT-5 tailored to different subscription levels, suggesting that only Pro subscribers may gain access to the advanced capabilities of o3 through GPT-5.
Introducing o3-mini: A New Approach
In the meantime, OpenAI has rolled out o3-mini in three distinct forms within ChatGPT. This streamlined version offers faster and more cost-effective functionality, making it a more practical option for users. As we look ahead, it seems unlikely that o4, the expected successor to o3, will be introduced in its original concept. Instead, GPT-5 and its successors are set to incorporate the advanced thinking abilities that were initially designated for the o models.