SP 3 (simplified P 3 ) theory is widely used in LWR (light water reactor) analyses to partly capture the transport effect, especially for pin-by-pin core analysis with pin size homogenization. In this paper, a SP 3 code named STELLA is developed and verified at SNERDI (Shanghai Nuclear Engineering Research and Design Institute). For SP 3 method, neutron transport equation can be transformed into two coupled equations in the same mathematical form as diffusion equation. In this work, SANM (semi-analytic nodal method) is used to solve diffusion-like equation, due to its easy to handle multi-group problem. Whole core nodal boundary net current coupling is used to improve convergence stability in SANM, instead of solving two-node problem. CMFD (coarse-mesh finite difference) acceleration method is employed for 0-th SP 3 equation, which represents the neutron balance relationship. Three benchmarks are used to verify the SP 3 code, STELLA. The first one is a self-defined one dimensional problem, which demonstrates SP 3 method is extremely accurate, due to no academic approximation in one dimensional for SP 3 . The second one is a two dimensional one-group problem cited from Larsen's paper, which is usually used to verify and prove the SP 3 code correct and accurate. And the third one is modified from 2D C5G7-MOX benchmark, whose numerical results indicate that STELLA is accurate and efficient in pin size level, compared to diffusion model.