Steel fiber reinforced concretes currently used do not have enough post-cracking tensile strength to be used alone, without main reinforcement (rebars), in a beam. On the other hand, replacement of the secondary reinforcement (stirrups) seems promising. The use of fibers would save considerable time and would facilitate placement in highly reinforced structures. A large experimental program was carried out to quantify the mechanical contribution of fibers in beams. In a first step, five large scale double-T-girders were tested, with the transverse reinforcement varied - concrete only, stirrups, or fibers. A global optical method, stereophotogrammetry, was used to monitor cracking. It showed a mainly mode I crack opening process. Furthermore, these tests suggested a scale effect linked to the height of the beam: the higher the height of the beam, the wider the crack opening at failure, i.e. the smaller the residual stress carried by the fibers. These findings lead to optimize both material and structure in regard to shear behavior. A high-strength steel fiber reinforced concrete was used in small rectangular beams, 0.25-m high, to provide high residual stress of the material and small crack opening at failure. For both types of beams--T-girders and rectangular beams--the mix was optimized using the Baron-Lesage method and the mechanical characterization of the material was conducted using a uniaxial tensile test monitored by the rate of the crack opening. A method of analysis based on a block mechanism is proposed and compares well with experiments.