Lots of models CAN handle large contexts, gemini 2.5 pro their latest model can take 1 million tokens of context