This study is part of the global effort to develop precise and accurate mechanical testing methodologies for metallic parts in as-built condition produced by powder-bed additive manufacturing processes. Miniature 316L stainless steel specimens with a 7 × 1 × 1.5 mm3 reduced section were produced by laser powder bed fusion. Three distinct surface conditions were considered. Standard specimens were also produced for comparison. Special attention was given to the non-destructive measurement of the cross-sectional area of small as-built specimens. The ultimate tensile strength of the as-built miniature specimens was 12% below standard-sized or machined specimens. Using caliper measurements lead to an overestimation of the section by as much as 20% compared to CT-scan values, resulting in an underestimation of UTS. Point micrometer measurements yielded more accurate estimations of cross-sectional area. With the experimental error taken into account, the specimen dimensions and surface condition influenced UTS. A coarser average microstructure for miniature as-built specimens was linked to a reduction of UTS by about 4% compared to standard specimens produced in the same conditions. Despite a larger scatter obtained with miniature specimens, no effect of specimen dimensions on the average value of elongation at break could be observed. Higher surface roughness of the order of Ra = 15 μm resulted in an elongation at break reduced from 52.7 ± 5.2 to 39.2 ± 5.0% for as-built miniature specimens. It was concluded that using large-scale or machined specimens may produce less accurate tensile characterization of thin 316L sections produced by laser powder bed fusion. Mechanical coupons with dimensions comparable with the structural part of interest should be employed. Accurate measurement of the cross-sectional area of small as-built specimens is critical for accurate measurements of mechanical strength. These results rationalize the scatter on tensile properties of LPBF 316L available in the literature.