Gemini 3.1 Pro's 1M token context window (far surpassing the 200K–256K of most open-source models) is largely thanks to Google's significant investment in computing power and storage infrastructure. In short, money talks. Who's still saying my expenses are too high?