Poly(benzodifurandione) (PBFDO) has emerged as a promising n-type mixed conductor for organic electrochemical transistors (OECTs), combining outstanding electrical properties with long-term stability. However, its intrinsically high electrical conductivity, advantageous for bioelectronics, leads to excessive operating currents and elevated energy demands in neuromorphic computing. To overcome this limitation, we introduce two complementary strategies that decouple intrinsic conductivity from computing performance. First, molecular-weight engineering via benzofuranone end-capping yields a reduced-chain-length polymer (PBFDO-BF) with substantially suppressed intrinsic conductivity. Second, ionic doping with 5 wt.% LiTFSI enhances ion-mediated conductance modulation in PBFDO-BF, enabling pronounced synaptic functionality. The combined PBFDO-BF + LiTFSI system achieves optimal OECT operation, characterized by enhanced modulation, pronounced hysteresis, and spike-dependent plasticity with long-term potentiation/depression (LTP/LTD), while reducing operating currents by approximately one order of magnitude compared to pristine PBFDO. In MNIST-based convolutional neural-network (CNN) simulations, this device delivers superior performance, attaining 97.78% training accuracy, 98.57% inference accuracy, and the lowest cumulative energy consumption to reach ≈90% accuracy. These findings establish molecular-weight control coupled with ionic doping as an effective design paradigm to optimize PBFDO for energy-efficient neuromorphic OECTs, without compromising stability or solution processability.