Vipi Radio
OpenAI has unveiled o1, a new “reasoning” model designed for faster and more complex problem-solving than its predecessors, alongside a smaller, more affordable version called o1-mini.
This model, also known as the highly anticipated Strawberry model, improves code writing and multistep problem-solving but comes with higher costs compared to GPT-4o.
The release of o1 is considered a “preview,” with access granted to ChatGPT Plus and Team users now, and to Enterprise and Edu users next week.
o1’s training involves a new optimization algorithm and dataset, and it uses reinforcement learning and a “chain of thought” approach for processing queries.
