Because they don't have a lot of parameters to store general Wikipedia knowledge. They're small. Use big models that have high parameter capacity to store general information. Or build a harness around the small model that searches a knowledge base/internet.
Use the right tool for the job. It's like asking why a screwdriver isn't good at sawing wood, or calling C a terrible language because it's hard to make CRUD apps with it.
I asked Gemma 4 E2B, and if you use it as a reasoning model, it will give a better answer (that it doesn't know; it was also using the date information from the prompt.)
