It has been established experimentally that both natural gas and hydrogen discharges, through circular orifices larger than a critical diameter, sustain stable lifted flames irrespective of the reservoir pressure driving the release. At smaller diameters, however, stable burning will only be achieved at operating pressures higher than a particular, diameter-dependent, threshold. This latter is strongly fuel-type dependent but empirical correlations have been developed to describe such behaviour. Given the wide disparity of critical diameters depending upon the fuel type considered, the behaviour of multi-component gaseous mixtures involving hydrogen is much less predictable than the correlations for pure fuels might indicate. A series of experiments has been undertaken in which H2-CO mixtures are discharged from a high pressure reservoir to ambient through convergent circular nozzles, varying in diameter from 1.3 to 5 mm. A wide range of driving pressures has been investigated, from 3 to 50 bar, embracing fuel mixtures containing up to 20% CO by volume. Stability curves have been derived that identify the region where stable burning is sustained relative to the pressure ratio (reservoir pressure/ambient pressure) and the CO concentration in the fuel mixture. The present experimental data are compared with existing correlations (Kalghatgi 1981), derived from subsonic releases, and Birch et al. (1988a), applied to underexpanded supersonic methane jet flames. Although these correlations reproduce the general trend observed experimentally, they appear to significantly over-estimate the stability region. Furthermore, they do not account for the high sensitivity of the diluent concentration on the blowout stability. A thorough investigation is carried out in order to determine the source of discrepancies observed between the empirical correlations and the present measurements. Differential diffusion between the fuel components is shown to be negligible. The introduction of additional information on the turbulent flowfield from numerical simulation of the isothermal underexpanded jet, immediately upstream of the flame stabilisation region, does not appear to yield discriminating evidence for incipient blow-off instability. The maximum burning velocity alone, a widely employed characterising parameter, does not appear to describe accurately the complex interactions between the turbulent flowfield and the chemical kinetics.