OpenAI is changing the path for one of its most anticipated AI model releases. The company no longer plans to launch o3 as a stand-alone model, according to CEO Sam Altman, and will instead fold that work into GPT-5.
The move is framed as a product simplification. Rather than asking users to choose from an expanding set of models, OpenAI says it wants ChatGPT and its API to feel more unified and easier to use.
What Changed With o3
o3 had been positioned as OpenAI's next major AI model. The company originally said in December that it wanted to release o3 sometime early this year, and Kevin Weil, OpenAI's chief product officer, recently said it was on track for a "Febr uary-March" launch.
That plan has now shifted. In a post on X on Wednesday, Altman said OpenAI intends to release GPT-5 in the coming months and that GPT-5 will integrate a lot of OpenAI's technology, including o3. Because of that decision, o3 is no longer planned as a separate release.
The practical message is simple: users who expected a distinct o3 model should now look toward GPT-5. OpenAI is presenting GPT-5 not just as another model name, but as a broader system that brings together model capabilities and ChatGPT tools.
Why OpenAI Wants Fewer Choices
Altman's explanation centered on complexity. OpenAI's model lineup has become harder for users to navigate, especially inside ChatGPT, where picking the right model can feel like a product decision rather than a simple conversation with AI.
Altman wrote that OpenAI wants to do a better job sharing its intended roadmap and simplifying its product offerings. He also said the company wants AI to "just work" for users and wants to move back toward "magic unified intelligence."
That framing matters because the decision is not only about o3. It is also about how OpenAI wants people to experience ChatGPT. The company appears to be moving away from a future where users must understand the differences between many models before they can get the best answer.
In plain terms, the goal is a system that can decide more of that for the user. According to Altman, OpenAI wants systems that can use all of its tools, know when to think for a long time or not, and be useful across a wide range of tasks.
How GPT-5 Access Is Expected To Work
Altman said OpenAI plans to offer unlimited chat access to GPT-5 at the "standard intelligence setting" once the model is generally available. That access will be subject to "abuse thresholds," though Altman did not provide more detail about what the setting or thresholds mean.
Subscribers will still have differentiated access. ChatGPT Plus subscribers will be able to run GPT-5 at a "higher level of intelligence," while ChatGPT Pro subscribers will be able to run it at an "even higher level of intelligence."
GPT-5 is also expected to incorporate several recent ChatGPT features. Altman specifically mentioned voice, canvas, search, deep research, and more.
- o3 will be integrated into GPT-5 rather than released by itself.
- GPT-5 is planned for ChatGPT and the API.
- Unlimited chat access is planned at the "standard intelligence setting," subject to "abuse thresholds."
- ChatGPT Plus and ChatGPT Pro users are expected to receive higher intelligence settings.
GPT-4.5 Comes First
Before GPT-5 arrives, OpenAI plans to release GPT-4.5 in the next several weeks. Altman said the model is code-named "Orion" and described it as the company's last "non-chain-of-thought model."
That detail points to a larger technical direction. o3 and other OpenAI reasoning models are designed differently from non-chain-of-thought models. According to the source article, non-chain-of-thought models tend to be less reliable in areas such as math and physics.
Reasoning models take a different approach. They effectively fact-check themselves, which can help them avoid some problems that normally trip up AI models. That process can add latency, with reasoning models taking seconds to minutes longer to reach solutions, but they tend to be more reliable and capable.
OpenAI's shift therefore has two layers. It is simplifying the product for users, and it is also leaning further into reasoning models as a core direction for future releases.
The Competitive Context
The timing also comes as competition in AI has intensified. Chinese AI lab DeepSeek recently drew major attention with its R1 model, which matched o1 on a number of benchmarks. Unlike o1, R1 is described as an "open" model under a permissive license, meaning developers can download it and use it as they see fit.
Altman has acknowledged in recent social media posts that DeepSeek has reduced OpenAI's technological lead in AI. He also said OpenAI would "pull up some releases" to compete more effectively.
There have also been reports about GPT-4.5, or Orion, facing performance-related challenges and technical setbacks. Bloomberg, The Information, and The Wall Street Journal have independently reported that Orion has shown less of an improvement over GPT-4o than GPT-4 did over GPT-3.
For users and developers, the near-term takeaway is that OpenAI's roadmap is being reorganized around fewer visible model choices and a more integrated GPT-5 release. o3 is still part of that future, but not as the separate product many expected.