What happened
Open reasoning models force teams to talk about cost and control in plain terms. Shops that once ran one premium model for everything now mix cheaper reasoning options, self-hosted runs, and routing rules that match model strength to task difficulty.
Why it matters
Reasoning jobs get expensive: long context, retries, tools. If an open model is good enough for part of the work, you can reserve pricey providers for cases where quality, reliability, or ecosystem support actually justify the bill.
Directory impact
Place this near DeepSeek, ChatGPT, evaluation tools, and skills that emphasize verification. Cost pressure does not mean “pick the cheapest model.” It means each step has a quality bar and a model choice that matches it.
What to watch next
Look past token price to routing, self-hosting maturity, and honest benchmarks. Total task cost matters: a cheap model that needs five retries can cost more than a stronger model that finishes once.