Assessing the uncertainties in the estimates obtained from forest carbon budget models used for national and international reporting is essential, but model evaluations are rarely conducted mainly because of lack of appropriate, independent ground plot data sets. Ecosystem carbon stock estimates for 696 ground plots from Canada's new National Forest Inventory enabled the assessment of carbon stocks predicted by the Carbon Budget Model of the Canadian Forest Sector 3 (CBM-CFS3). This model uses country-specific parameters, incorporates all five ecosystem carbon pools, and uses a simulation-based approach to predict ecosystem C stocks from forest inventory data to implement a Tier-3 (most complex) approach of the Intergovernmental Panel on Climate Change Good Practice Guidance for Land Use, Land-Use Change and Forestry (IPCC-GPG). The model is at the core of Canada's National Forest Carbon Monitoring, Accounting, and Reporting System (NFCMARS). The set of ground plots meets the IPCC-GPG standard for model evaluation as it is entirely independent of the model, but similar in type to that required for IPCC Tier-3 inventory-based C stock estimation. Model simulations for each ground plot used only the type of input data available to the NFCMARS for the national inventory report in 2010 and none of the model's default parameters were altered. Ecosystem total C stocks estimated by CBM-CFS3 were unbiased (mean difference=1.9Mgha−1, p=0.397), and significantly correlated (r=0.54, p=0.000) with ground plot-based estimates. Contribution to ecosystem total C stocks error from soil was large, and from deadwood and aboveground biomass small. Results for percent error in the aboveground biomass (7.5%) and IPCC defined deadwood (30.8%) pools compared favourably to the IPCC-GPG standards of 8% and 30%, respectively. Thus, we concluded that the CBM-CFS3 is reliable for reporting of C stocks in Canada's national greenhouse gas inventories. However, available standards for judging model reliability are few, and here we provide recommendations for the development of practical standards. Analyses by leading species (n=16) showed that error could often be attributed to a small subset of species and/or pools, allowing us to identify where improvements of input data and/or the model would most contribute to reducing uncertainties. This C stock comparison is one of the first ever to follow the evaluation process recommended by the IPCC-GPG for a Tier-3 model, and is a first step towards verification of greenhouse gas emission and removal estimates based on C stock changes.