New ChatGPT model designed to break down problems, ‘think’ before responding

Oprah Winfrey recently aired a special all about AI, titled “AI and the Future of Us.” After noting that “AI is still beyond our control,” Winfrey said humans will be able to adapt, saying, “The stakes could not be higher.”

The comments came with guests including OpenAI CEO Sam Altman, tech influencer Marques Brownlee and current FBI Director Christopher Wray. Winfrey noted that she was ready for what AI could do for humanity, as the new o1 model is released for ChatGPT.

Nicknamed “Strawberry,” the new model is designed to think before responding. With hype around the new model, there is already debate on whether it lives up to those abilities.

Living up to the hype may be important for Altman, as this model is four times more expensive than the previous model GPT-4o. OpenAI admits that the older version may still be a better option for most everyday tasks.

What sets the new model apart is the ability to break down big problems into smaller steps and “think” through each one. This new feature can be powerful for tackling complex questions, but it’s not always the right fit for simpler tasks.

One user tested ChatGPT o1 by asking for help planning a Thanksgiving dinner for 11 people. The model paused to think, then gave a detailed 750 word response. It even suggested renting a portable oven, showing that it can give advice for a complex scenario.

However, when given a simple question, it delivered an 800 word essay. Given the same question, GPT-4o answered in just a few sentences.

That is where cost comes into play. The new ChatGPT model charges for what are called “reasoning tokens,” the extra steps it takes when thinking through a problem, which can quickly add up. This makes it less practical for quick, straightforward questions.

Experts are intrigued by the model’s potential for solving complex issues through step-by-step reasoning, but even Altman is managing expectations. The new o1 model is now available, and Altman said they are planning to bring a miniature version to all ChatGPT users for free.

OpenAI said it plans to continue to develop new models in the GPT series.