OpenAI o1 is here, a new Strawberry AI model that thinks before responding

Remember that OpenAI’s project Strawberry, it is now officially called OpenAI o1. On September 12, the company fully unveiled the new AI model that thinks before responding. According to OpenAI, the model can reason through complex tasks and solve harder problems than previous models in science, coding, and math. While it is the first of this series in ChatGPT, OpenAI expects to add regular updates and improvements. But all of this comes at a higher price. Let’s deep dive into its working.
OpenAI’s o1 model marks a significant stride. According to the company, the model is a progressive step toward achieving human-like AI. It also demonstrates tangible improvements in key areas, such as generating code and tackling complex, multistep challenges with greater proficiency than its predecessors.
OpenAI, in the release, stated that it has trained these models to spend more time thinking through problems before it responds, much like a human would. Through training, it learned to refine their thinking process, try different strategies, and recognise their mistakes. Adding on, it says that the model excels in math and coding. In fact, in the upcoming update, the OpenAI o1 performs similarly to PhD students in challenging benchmark tasks in physics, chemistry, and biology.
OpenAI confirmed that the new o1 model lacks several features that can be found in the ChatGPT 4o model. As an early model, it doesn’t have features like browsing the web for information and uploading files and images. This means that, for now, it can only take text prompts. So, if you think you can just click a picture, upload it here and it solves the problem; you are ahead of time.
In short, the OpenAI o1 version will only help you solve complex math problems and other academic problems that require reasoning. But there is more.
OpenAI CEO Sam Altman posted on X, saying that o1 is not perfect, it is flawed and still limited. But why did he call the new model “flawed”? Read here in detail. He also shared graphs representing how the new OpenAI o1 performed in solving reasoning tasks like Math, Code, and PHD-level science questions.
It is also much slower that the current model. OpenAI reasons that since it is designed to “think before answering”, it takes its own sweet time to solve the problem. The company also states that for many common cases GPT-4o will be more capable in the near term.
The company has established that the o1 series excels at accurately generating and debugging complex code. To offer a more efficient solution for developers, OpenAI has also released OpenAI o1-mini, a faster and cheaper reasoning model that is particularly effective at coding.
As a smaller model, o1-mini is 80 per cent cheaper than o1-preview, making it a powerful, cost-effective model for applications that require reasoning but not broad world knowledge.
ChatGPT Plus and Team users will be able to access o1 models in ChatGPT starting today. Both o1-preview and o1-mini can be selected manually in the model picker, and at launch, weekly rate limits will be 30 messages for o1-preview and 50 for o1-mini. The company is working to increase those rates and enable ChatGPT to automatically choose the right model for a given prompt.
While OpenAI is planning to bring o1-mini access to all ChatGPT Free users, the OpenAI o1 comes at a higher price. In the API, o1-preview is $15 per 1 million input tokens, and $60 per 1 million output tokens. For comparison, GPT 4o price is set at $5 for 1 million input tokens and $2.50 for 1 million input token. For India, it means that where GPT 4o is available at around Rs 420 for 1 million input token, the users have to pay Rs 1,260 for the same to access OpenAI o1.
The ChatGPT Plus subscription is priced at $20 per month, which translates to around Rs 1,650 per month in India.