[starchild3001]

The reality is that cloud and AI model serving has become a race to the bottom. Top models are increasingly similar in capabilities, so price, not features, has become the main differentiator.

OpenAI’s key move was to put its thinking capability behind a router, making it automatic and optional. This allows the system to use more expensive reasoning only when needed, reducing average serving costs.

They're clearly doing the right thing -- the router approach, in principle, will lower their costs while maintaining roughly the same quality.

Of all things, this benefits OpenAI's bottomline. It isn't necessarily an attack on Antropic or Google Gemini.

It's just inference costs slowly becoming "too cheap to meter", while user experience vastly improving thanks to reduced latency for easier-to-answer prompts.

It's a win for OpenAI (lower cost). It's a win for chatgpt users (lower latency and better model for the free tier). And stronger competition for anyone else, which eventually benefits everyone.

FYI, you can get a glimpse of the current AI-model market share here: https://openrouter.ai/rankings?view=week