OpenAI: GPT-4.1

GPT-4.1 is what you reach for when context length actually matters. A million tokens is still rare in practice — and when you do need it, GPT-4.1 is the one that holds together. Strong coding, good reasoning, competitive pricing (~$2/Mtok in). The trade-off is character: it's agreeable where Sonnet is blunt, which means you sometimes have to ask twice for the disagreement you wanted the first time. Best for: dumping large codebases or long transcripts in and asking for the takeaway; quick production scripts where you want function-calling to just work; work that benefits from the OpenAI ecosystem (image gen in the same turn, voice via the same vendor). Avoid for: nuanced editorial writing (Sonnet has better taste); architecture disagreements (Opus pushes back harder); anything where you need the model to tell you "no, that's the wrong approach" without being coaxed into it. Pricing frame: at $2 in / $8 out per Mtok, a 5-person team at 200 msgs/day lands around $65/month. Among frontier models, this is the best cost per context-length-window.