AI models
Long chats
Every model reads a limited amount of text.
Small models on your device get a shorter set of instructions so more of the chat fits: a 2B model (Gemma 4 E2B counts as one) keeps the last 4,096 tokens, and a model GlirOS leads step by step through big jobs keeps 8,192. When a chat gets longer than that, GlirOS keeps a short running summary of the earlier messages for the model, and a line in the chat shows where the model's view begins. The chat itself is never shortened.
This guide is the same text GlirOS itself uses to answer questions about GlirOS.