**OpenAI Unveils GPT-4.5**
OpenAI has officially launched GPT-4.5.
On Thursday, the organization revealed the latest version of its large language model (LLM) in a research preview. As OpenAI CEO Sam Altman previously stated, GPT-4.5 will be the final model lacking chain-of-thought reasoning, indicating that all subsequent models will include enhanced reasoning abilities.
According to OpenAI’s white paper, initial testing suggests that GPT-4.5 provides a more intuitive user experience. “Its broader knowledge base, improved alignment with user intent, and heightened emotional intelligence render it ideal for tasks such as writing, programming, and addressing practical issues—with reduced hallucinations,” the report notes.
During a demonstration, OpenAI team members explained that while GPT-4.5 is not specifically built as a reasoning model, it still has advanced knowledge and can aid in problem-solving. In contrast to the “o” series, which is tailored for scientific and mathematical tasks, GPT-4.5 is designed as a general-use tool.
This release represents OpenAI’s first significant GPT update since the introduction of GPT-4o in May. Over the last year, the company has also launched reasoning models such as o1, o3-mini, and o3-mini-high. These models reportedly focus on utilizing real-time computational power (test-time compute) instead of depending exclusively on extensive training and large datasets.
### The Significance of GPT-4.5
The emergence of GPT-4.5 also underscores the ongoing delays concerning GPT-5. Persistent speculation indicates that OpenAI has faced diminishing returns in model enhancements, prompting inquiries about whether merely increasing computational capabilities and dataset size yields better outcomes.
This might explain why GPT-5, despite high expectations, has not yet been released. Altman has asserted that GPT-5 will exceed his own intelligence, while former CTO Mira Murati described it as possessing “Ph.D.-level intelligence.” Whether GPT-4.5 indicates that OpenAI is approaching these lofty goals or simply using time remains to be determined.
According to unnamed former OpenAI employees referenced by *Fortune*, GPT-4.5—internally dubbed “Orion”—was meant to be GPT-5 at first. However, it allegedly did not reach the performance improvements necessary to warrant the next-generation designation. Altman has remarked that GPT-5 might still be launched within a few months.
### How to Access GPT-4.5
Starting today, GPT-4.5 is available for testing—but only for a limited number of users. Currently, accessibility is restricted to ChatGPT Pro subscribers, who pay $200 monthly. ChatGPT Plus subscribers, who pay $20 per month, will obtain access next week.
This change in OpenAI’s subscription model has provoked dissatisfaction among Plus users, who were previously granted early access to new features. Some are now questioning the value of their subscription.
While OpenAI may intend to increase Pro subscriptions, this approach risks alienating Plus users, especially as rivals like xAI’s Grok 3, DeepSeek R1, and Anthropic’s Claude 3.7 Sonnet continue to gain popularity.
With GPT-4.5 potentially signaling the status of GPT-5’s development, the upcoming weeks will determine whether it fulfills expectations—or disappoints amid the buzz.