Context: Near-infrared reflectance spectroscopy (NIRS) is a tool that permits rapid and inexpensive prediction of the nutritional characteristics of forages consumed by ruminants. Aim: Our aim was to investigate the feasibility of developing a NIRS calibration to predict the nutritional value of the majority of grasses, legumes and forbs that are utilised for sheep and cattle production in southern Australia. Methods: More than 100 annual and perennial forage species were grown in replicated plots at two locations over a period of 3 years. Biomass was sampled every 3-6 weeks, dried, ground and scanned with a desktop NIRS machine (n = 4385). One-quarter of these samples were subjected to laboratory analysis for calibration development or validation. Key results: Despite the large variation in the taxonomy and maturity of the plants when sampled, we successfully developed broad calibrations that predicted key nutritional traits. We achieved excellent predictions for crude protein, with a ratio of standard error of performance: standard deviation (RPD) of 5.3, and standard error of cross validation (SECV) of 1.06%. Predictions of neutral detergent fibre were also excellent (RPD 4.3, SECV. 3.5%). For pepsin-cellulase DM digestibility and acid detergent fibre, predictions were very good (RPD 3.7, SECV 2.6% and RPD 3.9, SECV 2.1%). Predictions for organic matter were less reliable (RPD 2.2). We achieved very promising predictions of methane production during batch culture fermentation (RPD 3.1, SECV 3.5 mL/gDM). Predictions of ammonia and total volatile fatty acid concentrations in the post-fermentation substrate were poor. Conclusions: We found that the broad calibrations predicted the nutritional traits of annual grasses, annual legumes and forb species with greater accuracy than perennial grasses or legumes. This could be associated with the accuracy of the wet chemistry methods. As a general rule, separating taxonomically similar species into groups before the development of calibrations, did not lead to more accurate predictions. Implications: If more spatial and temporal diversity can be built in without a large reduction in accuracy, these broad NIRS calibrations represent a valuable tool for Australian researchers, feed testing agents and livestock producers, as they encompass nearly all of the species that appear in monocultures or mixed swards.