OpenAI cancels o3 Model to favor GPT-5

Another breaking news from OpenAI. OpenAI is canceling and re-releasing new AI models. Instead of releasing o3, their next giant model, they will develop a "unified" model, maybe called GPT-5. To know more about the o3 model take a look here.

GPT-5 will merge some of OpenAI's different AI technologies, OpenAI CEO Sam Altman explains. It will be integrated into ChatGPT and the OpenAI API. That means that they will not release o3 as a standalone model.

Altman explained how OpenAI plans to make things easier. They want AI "just works" for everyone. He admitted that their current model and services are too complex for a normal user to use or understand they have a model specific for mathematics, one for PhD level research, and many more but the user doesn't see the description of the model while selecting one they one get to know the name of the model and the rest is up to users research in choosing the model but they are planning to fix this. They even don't like the model selector in ChatGPT and plan to make it easier to use.

What to Expect from GPT-5

OpenAI will provide unlimited chat with GPT-5 in a "standard intelligence setting" to everyone which will be really cool. This move can be because of the recent AI developments around the globe as DeepSeek R1 has created a new type of trend for AI models. ChatGPT Plus members will get a "higher level of intelligence," and ChatGPT Pro members will get an "even higher level" which could be the ultimate model for them/ We don't yet know what those levels are, but it seems like different levels of performance just like we saw in OpenAI o3 mini.

GPT-5 will also include voice, canvas, search, research features, etc. The overall vision is to create something that can leverage all of OpenAI's tools, know when to go deep and be helpful for lots of different tasks. This can be a huge step towards a smarter, more powerful AI assistant.

GPT-4.5 "Orion" Coming First

OpenAI will release GPT-4.5, code-named "Orion," in the coming weeks before GPT-5 arrives. Altman calls this their last "non-chain-of-thought model." That is, as opposed to o3 and other "reasoning" models, Orion won't be nearly as skilled at such as math and physics. Reasoning models can better fact-check themselves and thus are more reliable.

Why the Switch? Competition and Challenges

OpenAI seems to be doubling down on competition reasoning models. DeepSeek, a Chinese AI research institute, recently released their R1 model, which is similar to OpenAI's o1 and even open-sourced. There have also been rumors that GPT-4.5 (Orion) has performance and technical problems. It's not a huge step from GPT-4o, according to reports. Maybe another reason for the change of direction. OpenAI probably doesn't want their next major release, GPT-5, to be anything less than a quantum leap ahead. If you wanna use Deepseek R1 in your machine locally then you can check here. Altman has said that DeepSeek is catching up.

Key Takeaways

It will be fascinating to observe whether OpenAI is opting to focus on one, single AI model such as GPT 5 will enhance the user experience. And how the intelligence levels for the different ChatGPT levels pan out after GPT-5 is launched. Artificial intelligence has some thrilling times ahead!