In a livestream in the present day, OpenAI lastly introduced the launch of its GPT-4.5 mannequin, however with a twist: For now, utilizing it requires a $200 per thirty days ChatGPT Professional subscription.
That’s as a result of the brand new massive language mannequin, or LLM, continues to be technically in a “analysis preview” state. That is all apparently shorthand for “please pay us for the privilege of being a beta tester.”
Jokes apart, the corporate is promising GPT-4.5 offers extra “pure dialog,” and performs higher in the case of programming, sample recognition, writing, and “fixing sensible issues.”
Word that GPT-4.5 isn’t meant as a reasoning mannequin, that means it received’t have the self-correcting or deep analysis capabilities of OpenAI o1, o3-mini, or the upcoming full model of o3. In a leaked doc, the corporate stated it’s not contemplating GPT-4.5 a “frontier” mannequin, however it’s OpenAI’s largest LLM but, which ought to make it best for on a regular basis queries, as the massive database of coaching materials ought to make for faster solutions.
Credit score: OpenAI
In inside exams, OpenAI says the brand new mannequin hallucinated about 24.7% much less regularly than GPT-4.0, and was about 34.3% extra correct. Round 57% of inside human testers reportedly most popular GPT-4.5 to 4.0.

Credit score: OpenAI
In different phrases, GPT-4.5 isn’t precisely on the slicing fringe of AI, but it surely ought to be an enchancment on different non-reasoning fashions, offering a pleasant center floor between cheaper, extra conventional LLMs and their regularly paywalled successors. The one main omissions appear to be multimodal options like AI Voice Mode and video enter, though OpenAI says its engaged on updating its person expertise, which appears to trace some model of those capabilities may come to the mannequin sooner or later.
That is sensible, on condition that, as soon as once more, that is technically an early launch, with OpenAI saying it’s “nonetheless exploring” GPT-4.5’s limits and the way folks can use it. To that finish, the LLM is about to increase to all paid ChatGPT plans over the subsequent two weeks, earlier than presumably reaching free customers as a alternative for GPT-4.0 as soon as it leaves its preview state.