OpenAI has officially launched GPT-4.5, the highly anticipated AI model code-named Orion. The announcement, made on Thursday, confirms that GPT-4.5 is OpenAI’s largest and most powerful model to date, trained with more computing power and data than any previous iteration. Despite its scale, OpenAI does not classify it as a frontier model.
Access and Availability
Subscribers to ChatGPT Pro, OpenAI’s $200-a-month plan, can access GPT-4.5 immediately as part of a research preview. Developers using OpenAI’s paid API tiers also gain access starting today. Meanwhile, users on ChatGPT Plus and ChatGPT Team can expect to use the model sometime next week.
Performance Insights
GPT-4.5 follows the traditional AI training approach of scaling up computing power and data in its pre-training phase. This method has historically resulted in significant improvements in performance across multiple domains, including mathematics, writing, and coding. OpenAI asserts that GPT-4.5 has “a deeper world knowledge” and “higher emotional intelligence” than its predecessors.
However, benchmark tests suggest that the benefits of scaling up may be leveling off. While GPT-4.5 outperforms many models on OpenAI’s SimpleQA benchmark for factual accuracy, it falls short compared to newer reasoning models from competitors such as DeepSeek and Anthropic. Additionally, on coding benchmarks like SWE-Bench Verified and SWE-Lancer, GPT-4.5 matches or slightly outperforms GPT-4o but does not surpass OpenAI’s deep research models or Anthropic’s Claude 3.7 Sonnet.
Costs and Challenges
Running GPT-4.5 is expensive. OpenAI charges developers $75 per million input tokens (approximately 750,000 words) and $150 per million output tokens. This is significantly higher than GPT-4o’s cost of $2.50 per million input tokens and $10 per million output tokens. Due to these high operational costs, OpenAI is still evaluating whether GPT-4.5 will remain available in its API long-term.
Emotional Intelligence and Creative Capabilities
One of GPT-4.5’s standout features is its ability to generate responses with a more natural and empathetic tone. In tests, it excelled in tasks requiring social awareness, such as providing supportive responses to users experiencing personal challenges. It also demonstrated strong creative abilities, successfully generating graphical representations in SVG format—something that competing models struggled with.
The Future of AI Scaling
The release of GPT-4.5 highlights a broader industry challenge: the limitations of traditional AI scaling. OpenAI co-founder Ilya Sutskever previously noted that the era of pre-training with massive datasets may be reaching its peak. In response, OpenAI and other industry leaders are shifting focus toward reasoning models, which prioritize logical processing over raw scale.
OpenAI plans to integrate its GPT and “o” series reasoning models, beginning with GPT-5 later this year. While GPT-4.5 may not claim the title of the most advanced AI model, it is a critical stepping stone toward the next frontier in artificial intelligence.
