It has been quite some time since Open AI, one of the biggest and well-known names in the A.I. realm, launched their most recent and incredible A.I. model, the ChatGPT 4o.
The model was a huge success. It was cheaper than the ChatGPT 4.0 model it replaced while being more reliable, accurate, and a lot faster. While it is without a doubt the most useful A.I. model to date from Open AI, especially paired with the advanced voice mode, Open AI has been working on something completely different, and it is looking pretty awesome.
Open AI, in a new announcement, launched their latest offering, the all-new Open AI o1 model. Where does the 1 come from? “We are resetting the counter,” they said. Open AI says the model is a completely different approach, and they are turning it into a new series of A.I models starting with the Open AI o1.
The biggest highlight of the o1 A.I. model is that it can actually think before it sends out a response, a lot more like a human would. The model can also reason and refine the response, ruling out potential inaccuracies and mistakes.
Open AI also mentioned that the model excels in use cases where reasoning is required, and things like programming, mathematics, and physics-based problems could greatly benefit from it. They also said that the reasoning and thought process of the new o1 model is comparable to that of a Ph.D. student.
They also mentioned that in the International Mathematics Olympiad, a math exam, the new Open AI o1 model gave a lot more correct answers—83 percent compared to the 13 percent of the ChatGPT 4o. Furthermore, Open AI also reported improvements on the safety side of things, saying since the new o1 model can think and reason before generating a response, it can also adhere to the safety guidelines a lot better than previous models like ChatGPT 4o. For reference, Open AI said that the new model scored an impressive 84 in their hardest jailbreak test, where the ChatGPT 4o scored only 22 points.
The new Open AI o1 model is looking pretty impressive and will be available for ChatGPT Plus users right from the launch. However, since this is an early model, they said it does not yet have the same capabilities of the bigger ChatGPT 4o model, and things like searching the web and file uploads are not available yet.
Open AI also announced the smaller o1 Mini model alongside the Open AI o1 Preview model that can provide the same benefits in applications and use cases where a broad knowledge is not required. For example, programming could still benefit from a smaller model.
Open AI also said that the model will be rolled out to free users, and a lot more is yet to come for the new Open AI o-named models.