The GNSS State Space Representation (SSR) technology is widely accepted to be the most versatile approach for real-time GNSS corrections. It is employed in numerous commercial and scientific PPP and PPP-RTK services from a large variety of service providers. Its main advantage over observation space representation (OSR) techniques (e.g. RTK or network RTK) is the intrinsic support for broadcast applications, delivering corrections to an unlimited number of users. Furthermore, SSR allows augmentation services to focus on different quality aspects, for example accuracy, convergence time, coverage area, supported signals, integrity or bandwidth consumption. Until today, there is no SSR data format that is flexible enough to accommodate the needs of these different services. Here, we present the Geo++ SSRZ format that was designed to be a general, highly flexible and bandwidth-optimized format to support the complete variety of SSR applications. It provides clock, orbit, bias, as well as ionosphere and troposphere corrections for all GNSS, frequencies, and signals. The resolution of the data fields is adjustable and the service operator can choose from a variety of different messages to transmit the required corrections. Atmospheric information can be transported with several different models as for example spherical harmonics, Chebyshev polynomials or sampled grid points - or a combination of those. These features allow for high-resolution as well as bandwidth-optimized SSR corrections. The format allows to balance the bandwidth consumption over time for communication channels where a low, constant signal rate is available (e.g. L-Band satellite transmission, terrestrial radio). The following techniques and concepts are applied: the separation of static metadata from the actual SSR correction data message stream, entropy-encoding of SSR parameters, transmission of consistent SSR states based on SSR parameters with different update rates, temporal offsets, and complementary grids. The Availability of the SSR service can be improved with SSRZ by splitting satellite information in multi-constellation satellite groups and applying adaptive parameter resolution in case of a sudden decrease of bandwidth or increased SSR value ranges. In this paper, we introduce the different concepts and present results of a bandwidth and performance analysis when SSRZ is used to deliver centimeter-level Network RTK-quality corrections. The overall bandwidth consumption is shown to be about 0.11 to 0.17 bps/satellite/station for a multi-GNSS and multi-frequency constellation, with the exact value depending mainly on network size.