- OpenAI chief Sam Altman revealed that GPT-5 is coming in the next few months.
- GPT-5 will be an "AI system" that will integrate the GPT-series and o3 reasoning model, and other ChatGPT tools.
- GPT-4.5 is set to release in the next few weeks and it will be the last non-reasoning LLM from OpenAI.
OpenAI CEO Sam Altman has finally spilled the beans and disclosed the roadmap for GPT-4.5 and GPT-5 release, after stating that he is not going to be smarter than GPT-5. Altman on X shared that OpenAI is going to release the GPT-4.5 model next, which was earlier reported to be “Orion“. It may be released in the next few weeks.
GPT-4.5 will succeed the GPT-4o model which was unveiled in May 2024. Altman says GPT 4.5 will be the “last non-chain-of-thought model” from OpenAI which means that it will be a traditional language model and not a reasoning model.
After that, Altman says GPT-5 will be released in the next few months. OpenAI is going to unify the GPT-series and o-series reasoning model to create the GPT-5 model. In that sense, GPT-5 will be more of an “AI system”, rather than a standalone AI model.
Altman says GPT-5 will integrate the recently-announced o3 reasoning model, and other ChatGPT tools including web search, Canvas, Advanced Voice Mode, Deep Research, etc. The upcoming GPT-5 system will know when to “think” for longer, when to use the traditional LLM to respond quickly, when to browse the web, and so on.
The most surprising part is that OpenAI will no longer release the full o3 model. It will be part of the GPT-5 system. OpenAI is making these changes to simplify its model picker and product offerings.
Finally, Altman states that GPT-5 will be available to free ChatGPT users. And not just that, free users will have unlimited chat access to GPT-5 at the “standard intelligence” setting, which I think is the “medium” or “low” reasoning effort for the full o3 model. The rate limit for free users will be subject to abuse thresholds.
As for ChatGPT Plus users, they will have unlimited access to GPT-5, but at a “higher level of intelligence”, most likely the reasoning effort set to “high” for the full o3 model. And ChatGPT Pro users who pay $200 per month can access GPT-5 at an “even higher level of intelligence”.
File Analysis Using o3-mini and Model Spec Update
Besides GPT 4.5 and GPT-5, OpenAI also announced that both o1 and o3-mini reasoning models now support file and image uploads. It means that you can use the reasoning models to analyze documents and images. And ChatGPT Plus users can send up to 50 messages per day to o3-mini-high.
OpenAI has also updated its Model Spec which defines how its AI models behave. OpenAI’s models will uphold “intellectual freedom” and no longer censor topics, “no matter how challenging or controversial a topic may be”.
The shift comes amid criticism that ChatGPT is “woke” and liberal-leaning as claimed by Elon Musk. It appears Altman-led OpenAI is making efforts to win favor with President Trump and accommodate conservative viewpoints.
