Drift-flux model is a simple and reliable tool for void fraction prediction in two-phase flows. This work examines its performance in gas-liquid flow of small bubbles (<1 mm) at low void fractions (<10 −1) that resembles bubbly flow in human bloodstream during Decompression Sickness and can be also found in other two-phase applications such as flow boiling in macro-channels. Drift-flux model predictions are compared with experimental data measured in co-current upward bubbly flow for varying gas/liquid flow properties. Water and blood simulant are used as test liquids, while bubble size is controlled using prescribed surfactant (SDS) concentrations. Homogeneous flow model, which is a sub-case of drift-flux model, predicts adequately void fraction in water and blood simulant when the mean bubble size is lower than 200 μm. For larger bubbles in water (mean size up to 800 μm) and in blood simulant (mean size up to 350 μm), the performance of thirteen drift-flux models from literature is examined. Ten of them concern specifically bubbly flow, whereas three models are applicable to a wide range of flow conditions. To our knowledge, this is the first time that drift-flux models are tested at combined low void fractions (between ∼10−3 and ∼10−1) and sub-millimeter bubble sizes, so the determination of drift-flux parameters (distribution parameter, C0 and drift velocity, Ugm) is elaborated. On this account, a bubble size dependent correlation is proposed for Ugm determination.