OpenAI has officially unveiled GPT-4.5, the latest iteration of its large language model (LLM).
This announcement underscores a significant milestone, as GPT-4.5 will be the last non-chain-of-thought model, marking a shift towards introducing reasoning capabilities in all future models. This was confirmed by OpenAI’s CEO, highlighting a new era in AI development.
According to initial findings from testing, interacting with GPT-4.5 is reported to be more seamless than previous versions. Its enhanced knowledge base, improved alignment with user intent, and better emotional intelligence position it as a versatile tool for a variety of tasks such as writing, programming, and problem-solving, all while reducing the occurrence of errors.
During demonstrations, the OpenAI team clarified that although GPT-4.5 is not classified as a reasoning model, it boasts advanced capabilities that can help users tackle complex problems. This version is designed for general use, unlike the previous “o” series, which focused on scientific and mathematical applications.
The launch of GPT-4.5 follows the introduction of GPT-4o last May, which represents OpenAI’s ongoing commitment to refining its model offerings. In the past year, OpenAI has also released reasoning models, including o1, o3-mini, and o3-mini-high, as part of its strategy to prioritize real-time model performance over extensive training methods.
Moreover, the debut of GPT-4.5 highlights the ongoing anticipation surrounding the development of GPT-5. Despite the high expectations set for this upcoming model, delays have raised questions about the feasibility of significant improvements in performance. Reports suggest that OpenAI has been grappling with the challenges of scaling laws in AI model training.
Insiders indicated that GPT-4.5, internally known as Orion, was initially intended to be the flagship GPT-5 but did not meet the performance benchmarks necessary for that title. The release of GPT-5, according to OpenAI’s leadership, is expected in the coming months.
How to Access GPT-4.5
Starting today, users can begin testing GPT-4.5; however, access is currently limited to ChatGPT Pro members who pay $200 monthly. ChatGPT Plus subscribers will gain access next week for a monthly fee of $20. This staggered release has sparked questions among Plus subscribers, who were previously accustomed to early access to new features.
As OpenAI aims to boost its Pro subscriptions, there are concerns that this approach may alienate Plus users, creating an opportunity for competing AI tools to gain traction. Alternatives like Grok 3, DeepSeek R1, and Claude 3.7 from other companies are increasingly emerging as formidable options.
The anticipation surrounding GPT-4.5 is considerable, as its performance could provide insights into the progress of the highly anticipated GPT-5.
