Fanar 2 27B Instruct is a large open-weight model from QCRI built on the Gemma family architecture. It handles text-in, text-out tasks with a generous 131K token context window, making it capable of processing long documents in a single pass. As an instruct-tuned model, it's oriented toward following directions and responding to queries rather than raw completion tasks.