Sediments from the Upper Columbia River, Washington, USA, are contaminated with metals from smelting operations. We conducted short‐term and long‐term tests with the midge Chironomus dilutus and the amphipod Hyalella azteca and short‐term tests with the freshwater mussel Lampsilis siliquoidea with 54 sediments from the Upper Columbia River to characterize thresholds for toxicity of metals to benthic invertebrates. Test sediments were screened for toxicity by comparisons with low‐metal reference sediments. Toxic effects on amphipods occurred primarily in sediments from the upstream (riverine) reach, and toxic effects on midges occurred in sediments from both the upstream reach and the downstream (reservoir) reach. Little toxicity was observed in mussel tests. Toxicity thresholds (20% effect concentrations [EC20s]) for metals in sediment and porewater were estimated from logistic concentration–response models. Copper (Cu) concentrations in the simultaneously extracted metal fraction of sediments and bioavailable Cu in porewater, as characterized by biotic ligand models, had consistent associations with toxicity endpoints. Concentration–response models for sediment Cu produced EC20s for 6 endpoints, with long‐term amphipod survival and reproduction being the most sensitive. A logistic regression model fitted to an endpoint sensitivity distribution for sediment Cu predicted that approximately one‐half of the sediments tested would be toxic to at least one endpoint and that approximately 20% of test sediments would be toxic to more than half of the endpoints. These results indicate that sediments from the upstream reach of the Upper Columbia River, which contain high concentrations of metals associated with slags, cause a wide range of toxic effects in laboratory tests and are likely to have adverse effects on benthic invertebrate communities.