OpenAI has launched GPT-4.5, its latest AI model, codenamed Orion.
While it’s the company’s largest model yet, OpenAI does not consider it a frontier model, a term for the most advanced AI systems.
GPT-4.5 is now available for ChatGPT Pro users ($200/month) and developers using OpenAI’s API.
Other users, including those on ChatGPT Plus and ChatGPT Team, will get access next week.
The model follows the same training approach as previous versions, scaling up computing power and data to boost performance.
This method has led to major improvements in maths, coding, and writing in the past.
However, results from GPT-4.5 suggest that simply increasing scale may not be enough to drive big leaps forward anymore.
While it performs well on factual Q&A tests and makes fewer mistakes than earlier models, it lags behind some competitors, including Anthropic’s Claude 3.7 Sonnet and DeepSeek’s R1, in reasoning tasks.
It also lacks some features, like ChatGPT’s two-way voice mode, and is expensive to run, meaning OpenAI may not keep it available in its API long-term.
GPT-4.5 outperforms GPT-4o on some tests, particularly in factual accuracy and creative tasks, but struggles against newer AI reasoning models.
OpenAI says that standard AI benchmarks don’t always reflect real-world use, so it’s releasing GPT-4.5 as a research preview to better understand its strengths and weaknesses.
Looking ahead, OpenAI plans to merge its general AI models with reasoning models in GPT-5, expected later this year.
GPT 4.5 In brief:
GPT-4.5 is OpenAI’s biggest model yet, but it’s not classed as a frontier AI system.
Performance is mixed—it’s strong in factual accuracy but less competitive in reasoning tasks.
GPT-5 is on the way, aiming to combine general AI with reasoning capabilities.
This update isn’t a major breakthrough, but it’s a step toward OpenAI’s next big development.
Basically, the AI version of “trust the process.”