A new data model for chemical kinetics is presented, which clarifies the chemistry knowledge and assumptions that underlie large chemical kinetic models. With the new data model, it is feasible to remove most of the chemistry information and hidden assumptions typically hard-coded in automated mechanism-generation software, and bring them out into a human-readable database. This has significant practical advantages, including simplified maintenance of the software as chemistry knowledge improves. In the new data model, chemistry information is organized along intuitive functional-group lines. The new data model should facilitate peer-review and comparisons between even very large combustion chemistry models, and clarify where there are real differences in what is assumed about the chemistry. New model-construction, model-reduction, and least-squares fitting algorithms are presented. Applications relevant to design of new engines and fuels will be presented.