Sample requirements vary according to sample and analysis type.
For regular radiocarbon analysis of samples as solid graphite, the material submitted is processed to CO2 and subsequently reduced to solid carbon on an iron catalyst (graphite). In this case, the process blank and counting statistics limit the precision. The ideal sample size is 1 milligram of pure extracted carbon, though we routinely analyze much less than this (0.020 - 1 mg C). The reported error scales up with diminishing mass and age because the older the sample, the fewer the number of 14C counts in the accelerator and when the sample is small, the effect of the addition of exogenous carbon from the process blank is more pronounced.
For the MICADAS gas interface system (GIS), the sample masses should be in the size range of 100-300 µg C and while the process blank is much smaller for GIS than it is for samples processed to graphite, there are other sources of error (such as gas trapping) that cause the method to have higher overall variability. Currently the precision is approximately ± 10 ‰ for modern carbonate samples (vs ± 2 ‰ for graphite). For a carbonate sample that is 5,568 radiocarbon years (Fm = 0.5) the estimated precision would be ± 96 y (vs ± 24 y for graphite) and for a sample that is 18,497 yrs an age error of ± 200 y (vs ± 95 y for graphite). The former comparisons are for regular size graphite (1 mg C).
Considerations regarding which method is suitable for your sample sizes and scientific question include cost, expected error, sample mass and batch size. MICADAS Gas Interface samples are currently $134/sample for US Federal and Academic customers (vs $268/sample for graphite) and must be submitted in batches of at least 20 samples.
For more information see General Sample Requirements for regular AMS analysis and Sample Requirements for the MICADAS Gas Interface.
For more information regarding the potential for 14C-contamination of samples see here.