A larger number of qubits allows us to do effective quantum error correction. The idea is to group multiple physical qubits into one logical qubit, think of it as redundancy.
So what's the number of logical qubits we have achieved working practically then? Is this scalable, or is it just going to exponentially require physical qubits for each additional logical qubit?
Quantum error correction has been experimentally demonstrated for a single logical qubit, e.g. [0][1]. Even though there might be implementations of multiple such qubits, we're still very much in the "Noisy Intermediate-Scale Quantum" era.
Generally, the number of physical qubits scales linearly with the number of logical qubits.