Welcome to Neural. AI is evolving rapidly. We help you stay informed. Recently, Elon Musk initiated a bid to acquire OpenAI, leading Sam Altman to humorously rebuff the idea while teasing a potential Twitter purchase. While that was entertaining, the AI landscape appears to be settling down — WAIT, WHAT?! OpenAI has just unveiled plans for ChatGPT-5, set to launch this year?
Sam Altman, the CEO of OpenAI, has effectively reclaimed the spotlight with his announcement detailing the future of ChatGPT. He shared an OpenAI roadmap update that covers both the GPT-4.5 and GPT-5 AI models.
Altman begins by addressing the intricacies present within the current ChatGPT experience.
We aim to improve how we share our roadmap and significantly simplify our product offerings. Our goal is for AI to “just work” for you; we recognize how convoluted our model and product offerings have become.
The rapid development of new products is largely responsible for the complexity surrounding ChatGPT. Currently, ChatGPT Plus subscribers have access to the following models and their descriptions:
- GPT-4o: Ideal for most inquiries
- GPT-4o with scheduled tasks: Request follow-ups from ChatGPT
- o1: Employs advanced reasoning
- o3-mini: Quick at advanced reasoning
- o3-mini-high: Excellent for coding and logic
- GPT-4o mini: Quick for most questions
- GPT-4: Legacy model
Users face the dilemma of deciding whether to start with a large language GPT model or a reasoning o-series model. OpenAI’s challenge lies in determining which model will yield the most relevant responses. This is an area they are currently focusing on.
We feel the frustration with the model picker just as you do and aspire to return to a more seamless unified intelligence. Our next release will be GPT-4.5, the internally referred to model Orion, which will be our final non-chain-of-thought model. Following this, a key goal is to unify the o-series and GPT-series models to develop systems that utilize all our tools effectively, know when to engage in long-form reasoning, and be generally useful for a wide array of tasks.
In his update, Altman confirmed that the Orion model will be recognized as GPT-4.5, which is on track for release in the coming weeks. While it won’t feature the DeepSeek-like chain-of-thought functionality present in o3, ChatGPT-5 and later versions will incorporate this to demonstrate how the model processes information.
ChatGPT-5, according to Altman, will serve as a cohesive model that integrates both LLM and reasoning features, and chat capabilities will remain free of charge. He also noted that ChatGPT-5 is expected to launch in a few months.
For both ChatGPT and our API, GPT-5 will be released as a comprehensive system that combines much of our technology, including o3. We will discontinue o3 as a standalone model. The free version of ChatGPT will provide unlimited chat access to GPT-5 at the standard intelligence setting (!!), with certain safeguards against abuse.
In conclusion, Altman pointed out that GPT-5 would present a higher intelligence tier for paid subscribers, likely suggesting enhanced resource-intensive capabilities.
Plus subscribers will have the opportunity to operate GPT-5 at an elevated intelligence level, while Pro subscribers will access an even higher tier. These models will feature voice capabilities, canvas tools, search functionalities, deep research, and more.
One aspect of Altman’s announcement that stood out to me was the lack of lowercase formatting. Quite unusual, if you ask me.
In a previous update, Altman revealed that OpenAI aims to offer 10 deep research queries per month to the $20/month tier, while free users will start with two deep research queries. Currently, accessing deep research requires OpenAI’s $200/month ChatGPT Pro plan.
Definitely no lowercase letters there.
Stay tuned for more updates on the latest AI developments in the upcoming edition of Neural — exclusively on DMN! You can find the previous issue here.
