OpenAI unveiled a new AI model that thinks before responding. This new project is known as OpenAI o1 and initially we knew it by the name of project Strawberry.
This model excels at solving complex tasks, showing significant improvements in science, coding, and math. It also outperforms previous models in code generation and handling intricate, multistep problems.
OpenAI explained in its release that the models have been trained to approach problems more thoughtfully, similar to human reasoning. The training process involved refining their problem-solving methods, experimenting with various strategies, and recognising mistakes.
In the upcoming update, OpenAI o1 is expected to perform similarly to PhD students in challenging benchmark tasks in physics, chemistry, and biology.
OpenAI revealed that the new o1 model is missing several features present in the ChatGPT 4o model. As an earlier version, it cannot browse the web for information or handle file and image uploads.
OpenAI CEO Sam Altman posted on X, saying that o1 is not perfect, it is flawed and still limited.
It is also significantly slower than the current model. OpenAI explains that this is because it is designed to think before answering, which means it takes more time to work through the problem.
The OpenAI o1-mini is 80% more affordable than o1-preview, offering a cost-effective solution for tasks requiring reasoning without the need for extensive world knowledge.
In the API, o1-preview costs $15 per million input tokens and $60 per million output tokens. In contrast, GPT 4o is priced at $5 per million input tokens and $2.50 per million output tokens.