I’m trying to understand what to expect when a DeepSeek chat gets close to its context window. Does the interface warn you, stop accepting messages, or quietly lose access to older parts of the conversation?

I’ve seen different behavior across AI tools, and it’s hard to tell whether a forgotten detail means the limit was reached or the model simply missed it. If you’ve tested this with DeepSeek, what did you observe, and does the answer differ between the app and API?