Models that think before they answer: where reasoning earns its cost
18 September 2024

Models that think before they answer: where reasoning earns its cost

OpenAI has released a new kind of model that works through a problem step by step before answering.

It is slower and costs more. For the right work, it is worth it.

What was released

On 12 September OpenAI introduced o1, a model designed to reason. Instead of replying at once, it works through the steps of a problem first. It takes longer and costs more to run, and it handles harder questions noticeably better.

Where it earns its cost

  • Planning. Scheduling jobs with many constraints.
  • Checking. Comparing documents and finding what does not match.
  • Analysis. Working out why a number changed.
  • Complex requests. Cases with several conditions to weigh.

Where it does not

For simple questions, a faster and cheaper model does the job. Asking a reasoning model for a store's opening hours is a waste. The skill is in matching the model to the task: quick models for quick work, careful models for careful work. A well-designed system uses both.

Frequently Asked Questions

An AI model that works through a problem in steps before answering, which improves results on complex tasks.
No. Use them for multi-step problems. Simple tasks are better served by faster, cheaper models.

Find the multi-step work in your business that AI can now handle.

Tell us about your business and we will show you where AI agents can take work off your team.

Talk to us