We introduce a new framework that facilitates term structure modeling with both positive interest rates and flexible time-series dynamics but that is also tractable, meaning amenable to quick and robust estimation. Using both simulations and U.S. historical data, we compare our approach with benchmark Gaussian, stochastic volatility, and shadow rate models, where the latter enforces positive interest rates. Our approach, which remains arbitrarily close to arbitrage free, offers a more accurate characterization of bond Sharpe ratios due to a better fit of the volatility dynamics and a more efficient estimation of the return dynamics. Further, standard shadow rate and stochastic volatility models exhibit important restrictions that are largely absent in our approach.