Background Novel theranostic options for high-risk non-muscle invasive bladder cancer are urgently needed. This requires a thorough evaluation of experimental approaches in animal models best possibly reflecting human disease before entering clinical studies. Although several bladder cancer xenograft models were used in the literature, the establishment of an orthotopic bladder cancer model in mice remains challenging. Methods Luciferase-transduced UM-UC-3LUCK1 bladder cancer cells were instilled transurethrally via 24G permanent venous catheters into athymic NMRI and BALB/c nude mice as well as into SCID-beige mice. Besides the mouse strain, the pretreatment of the bladder wall (trypsin or poly-L-lysine), tumor cell count (0.5 × 106–5.0 × 106) and tumor cell dwell time in the murine bladder (30 min – 2 h) were varied. Tumors were morphologically and functionally visualized using bioluminescence imaging (BLI), magnetic resonance imaging (MRI), and positron emission tomography (PET). Results Immunodeficiency of the mouse strains was the most important factor influencing cancer cell engraftment, whereas modifying cell count and instillation time allowed fine-tuning of the BLI signal start and duration – both representing the possible treatment period for the evaluation of new therapeutics. Best orthotopic tumor growth was achieved by transurethral instillation of 1.0 × 106 UM-UC-3LUCK1 bladder cancer cells into SCID-beige mice for 2 h after bladder pretreatment with poly-L-lysine. A pilot PET experiment using 68Ga-cetuximab as transurethrally administered radiotracer revealed functional expression of epidermal growth factor receptor as representative molecular characteristic of engrafted cancer cells in the bladder. Conclusions With the optimized protocol in SCID-beige mice an applicable and reliable model of high-risk non-muscle invasive bladder cancer for the development of novel theranostic approaches was established.