Biological nitrogen fixation by the legume-rhizobia symbiosis is one of the most important mechanisms of nitrogen supply to plants, and depends on the symbiotic efficiency of the rhizobial strain and native rhizobia biodiversity is fundamental for the selection of new, more efficient strains for inoculant production. Soil samples were collected in areas with pastures, “sabiá” woodlots and Atlantic forest. Calopo plantlets were pre-germinated, transplanted to Leonard jars, inoculated with soil and harvested after 50 days. Four to five nodules per plant were used for isolation, from which 1575 isolates were obtained, distributed in 398 groups. Authentication was conducted for a single isolate from each group, followed by a symbiotic efficiency test.  Treatments were 122 isolates selected after authentication, based on shoot dry mass, five uninoculated controls with 0, 50, 100, 150 and 200 kg.ha-1 of N, and a control inoculated with SEMIA 6152, recommended for the culture, from which 25 isolates were considered efficient and selected for the unsterile soil evaluation, in which three cuts were performed at 45 days intervals.  Over 98% of the isolates have fast growth, although the recommended strain is Bradyrhizobium japonicum. After grouping, 163, 257 and 77 groups were found for the pasture, sabiá woodlots and Atlantic Forest areas, respectively, showing high diversity, but no significant effect of vegetation cover on rhizobial diversity. Plants inoculated with the most efficient strains did not differ from the treatment receiving 150 kg.ha^-1 of N. The highest proportion of efficient isolates was from sabiá woodlots, indicating that vegetation cover may affect symbiotic efficiency.  There was high variability in nodule dry mass, N content and accumulation, indicating large differences in symbiotic capability. Higher responses to the isolates were found in the second cut, and some isolates allowed shoot dry matter similar to those of the highest N fertilization treatments