Back in February, OpenAI CEO Sam Altman revealed the future roadmap for its AI models. He confirmed that GPT-4.5 would be the last non-chain-of-thought model and that GPT-5 would integrate the capabilities of both non-chain-of-thought and reasoning models like o3. Altman also mentioned that they would not be releasing o3 as a standalone model as initially expected.
Today, Sam Altman announced changes to these plans. OpenAI will be releasing standalone o3 and o4-mini models in a couple of weeks, followed by the GPT-5 model in a few months.
OpenAI is making these changes for several reasons. First, they found it more challenging than anticipated to smoothly integrate both reasoning and non-reasoning model capabilities. Second, they realized they could significantly improve GPT-5 beyond their original expectations. Third, they want to ensure they have sufficient compute capacity to support the unprecedented demand for GPT-5.
In a follow-up tweet, Altman mentioned that they had made substantial improvements to the o3 model, which would surely please customers. He also confirmed that the o3 model would feature a pro mode, enabling it to use more compute power for delivering the best responses to the most complex questions.
For general users, the free tier of ChatGPT will offer unlimited access to the forthcoming GPT-5 model at the standard intelligence level. Plus paid tier subscribers will have the option to run GPT-5 at a higher intelligence setting, while Pro paid tier subscribers can utilize GPT-5 at an even more advanced level of intelligence.
These updates indicate OpenAI's commitment to adjusting its roadmap and ensuring optimal user and developer experience. The phased model release strategy aims to address their internal technical hurdles while meeting the demand for advanced AI models.
0 Comments - Add comment