All AI chatbots can use conversation history as context. There is only so much a model can “remember.” That limit is the context length or context window.
Each model can only process so much of a conversation before it starts to “forget” earlier turns. Context windows are measured in tokens. A longer window lets the model keep more of the chat, attached files, and instructions in play.
For example, a 1,000,000-token window is roughly 750,000–900,000 words. If a chat grows past that window, the model may drop earlier parts of the exchange and give less relevant answers.
Magai shows when context is being cut off so you are not surprised. Context limits and usage multipliers differ by model and change as models are added or updated.
For the current list of Magai models, context windows, and multipliers, see Which AI models does Magai have access to? Magai currently includes 74 models. The default chat model is Claude Sonnet 5 at 0.8x.
If you are having long, in-depth conversations, pick a model with a larger context window from that list rather than relying on a copied table here.