Who Is OpenAI's O1 For?
It's more expensive, can handle edge case (albeit embarassing) questions, but still makes errors. Where would you use this in production that you wouldn't use 4o, vanilla or fine-tuned?
For who and what is this for?
It's better at planning a system design and asking the right questions (e.g. "Potential problems and challenges") then a 2-3 year experienced developer I would say. Will be very useful for scientific and biotech companies with less strengths in programming, but more in the science of their domain.
Basically, in order to stay relevant, Open AI and others are just going to do more meta trainings of smaller LLMs over LLMS, so you are going to get some specialized LLMS that do good for certain things, but you aren't going to see any massive improvement in functionality until someone fundamentally rethinks architecture down to the core.
o1 could be useful for heavy math/science/coding work, such as a study tool or coding assistant. If new o1-based tools come out and the model gets too good at this kind of work, many software engineers might be cooked. The codeforces result is concerning, although codeforces is not representative of software engineering in industry.
For me, it's substantially better at coding tasks than 4o.
It’s for VCs to go ooh and aah and boost their valuation.