The paper introduces a quantum computation model using local fermionic modes (LFMs), which are sites that can be either empty or occupied by a fermion. The authors show that one fermionic gate can be simulated by $O(m)$ qubit gates, and vice versa, where $m$ is the number of LFM sites. They demonstrate that the simulation cost can be reduced to $O(\log m)$ and a constant using different encodings. Nearest-neighbors fermionic gates on a graph of bounded degree can be simulated at a constant cost. A universal set of fermionic gates is identified, and the computation with Majorana fermions, which are halves of LFM sites, is studied. The paper also explores connections to qubit quantum codes and discusses the possibility of using fermions as decoherence-free quantum memory.The paper introduces a quantum computation model using local fermionic modes (LFMs), which are sites that can be either empty or occupied by a fermion. The authors show that one fermionic gate can be simulated by $O(m)$ qubit gates, and vice versa, where $m$ is the number of LFM sites. They demonstrate that the simulation cost can be reduced to $O(\log m)$ and a constant using different encodings. Nearest-neighbors fermionic gates on a graph of bounded degree can be simulated at a constant cost. A universal set of fermionic gates is identified, and the computation with Majorana fermions, which are halves of LFM sites, is studied. The paper also explores connections to qubit quantum codes and discusses the possibility of using fermions as decoherence-free quantum memory.