The genuine escape hatch. Ollama downloads an open-weight model and runs it on your own hardware, so there is no sign-up, no subscription, no usage limit and nothing sent to anybody. Pair it with a front end such as Open WebUI and you have a private ChatGPT-shaped interface that works on a plane. For people leaving over privacy, cost or the principle of the thing, this is the only option on the list that fully delivers.
The honest limitation is capability. A model that fits on a laptop is clearly weaker than a frontier model in a datacentre, especially on long reasoning and code, and you need a reasonably modern machine with plenty of memory before it is pleasant. Use it for the work that must stay private and keep a hosted assistant for the rest.