OpenAI launched GPT-4.5, also known as Orion, its largest and most powerful AI model yet. Despite its impressive scale, OpenAI does not consider GPT-4.5 a frontier model. The model is now available for ChatGPT Pro subscribers and developers on paid tiers of OpenAI’s API, with other users gaining access soon.
GPT-4.5 was trained with more computing power and data than previous models, offering improved performance in areas like writing, coding, and mathematics. It has a deeper world knowledge and higher emotional intelligence, although its performance on AI benchmarks, like reasoning tasks, still lags behind newer models from competitors such as DeepSeek and Anthropic.
Despite its size, GPT-4.5 is expensive to run. OpenAI charges developers $75 per million input tokens and $150 per million output tokens for API usage. Although the model has strengths, such as improved factual accuracy and emotional intelligence, it doesn’t surpass the capabilities of leading reasoning models in difficult academic tasks.
GPT-4.5 has demonstrated an ability to understand human intent more effectively, making it excel in creative tasks and social appropriateness. OpenAI claims the model is a step toward combining GPT models with reasoning models for better performance in real-world tasks.
OpenAI acknowledges the limitations of pre-training scaling laws and anticipates that the next evolution of its models will involve combining GPT with reasoning models, starting with GPT-5. Despite its challenges, GPT-4.5 is seen as a stepping stone toward even more powerful AI.
Highlights:
- GPT-4.5 is OpenAI’s largest model yet, trained with more computing power and data.
- Available to ChatGPT Pro subscribers and developers, with general availability soon.
- It excels in emotional intelligence, factual accuracy, and creative tasks.
- Still lags behind some competitors in academic and reasoning benchmarks.
- Very costly to run, with OpenAI exploring long-term API availability.
- OpenAI views GPT-4.5 as a step toward more advanced models combining reasoning and language capabilities.
