OpenAI gives users access to GPT-4.5 — for whom?

OpenAI has already started providing access to its newest Artificial Intelligence model, GPT-4.5, to users who have ChatGPT Plus subscriptions. The rollout to all subscribers is expected within three days.

TechCrunch reported about it.

Everything we know about the provision of access to GPT-4.5

The company reported on X that it expects the tariff limits to change. Initially, GPT-4.5 was launched for ChatGPT Pro subscribers for USD 200 per month.

"We would like to provide everyone who has access to GPT-4.5 with a significant speed limit, but we expect speed limits to change as we learn more about demand," the company wrote in its statement.

GPT-4.5 is now OpenAI’s largest AI model, training with more computing power and data than any of the company’s previous versions. Despite this, it is not the best OpenAI model. In some benchmarks, GPT-4.5 falls short of models from DeepSeek, Anthropic, and OpenAI itself.

GPT-4.5 is also very expensive to use, which even OpenAI acknowledges when considering whether to continue to serve the model in its API in the long run. To cover the costs, the company charges USD 75 per million tokens (about 750,000 words) entered into the model and USD 150 per million tokens generated by the model, which is 30 times the input cost and 15 times the output cost of the GPT-4o model.

OpenAI claims that the increased size of GPT-4.5 has given it a deeper knowledge of the world and higher emotional intelligence. According to the company, the model hallucinates less frequently than most models, which in theory could mean that it is less prone to fabrications.

The newest model is also a skilled rhetorician, as it proved in one of the internal tests, where it convinced another AI to give it cash and name a secret code word.

As a reminder, OpenAI updated the ChatGPT application for iPhone. Now, users can use the chatbot as a standard search engine by activating a special extension in the Safari browser.

We also wrote that OpenAI announced its plans to develop language models by talking about GPT-5. It will be the last model without chain-of-thought support.