Some of the thinking models might recover... with an extra 4k tokens used up in <thinking>. And even if they were stable at long contexts, the speed drops massively. You just can't win with this architecture lol.
Everything happening in the LLM space is so close to how humans think naturally.