Cerebrospinal fluid (CSF) sampling at the lumbar subarachnoid space (LSS) is technically challenging to learn. Currently, training relies on cadaver availability or performance in a clinical scenario. This study aims to develop and validate a low-cost, high-fidelity simulator to train in this technique. Using three-dimensional printing technology, a model of a healthy adult dog's lumbosacral vertebral column was produced. The model was augmented with synthetic materials and a fluidic system to replicate all procedural steps and permit successful CSF collection. The simulator was validated by four experts, who rated it highly across multiple criteria. Final-year veterinary students were recruited to take part in practical sessions using either the simulator (n = 16) or a cadaver (n = 16). Performance was recorded for each student, and feedback was obtained using an anonymous online survey. Student performance was similar between groups (p = .2), with 87.5% and 68.75% of students in the simulator and cadaver group, respectively, successfully placing the needle into the LSS. All successful students in the simulator group were able to obtain a CSF sample, versus none in the cadaver group. No difference in the number of attempts was detected between groups (p > .99), with most students taking more than three attempts. User experience was similar between groups, with 93.8% of students in each group rating the session as a positive learning experience. In summary, we demonstrate the validity of a novel, low-cost, and anatomically precise simulator that can be used for teaching CSF sampling at the LSS.