Frequent use of herbicides may impose a risk on non-target species. The objective was to test the combined toxic effect of binary herbicide mixtures—metribuzin:halosulfuron and metribuzin:flumioxazin—on non-target earthworms in two test systems: filter paper and a soil toxicity test system. The joint action experiments were independently run twice to substantiate the findings. The most potent individual herbicide was metribuzin, with a 50% lethal concentration (LC50) of 17.17 µg ai. cm−2 at 48 h in the filter paper test. The toxicity of the individual herbicides on the filter paper test was ranked as metribuzin>halosulfuron>flumioxazin. In the soil test, metribuzin and halosulfuron had high toxicity with an LC50 of 8.48 and 10.08 mg ai. kg−1, respectively, on day 14. Thus, the individual herbicide ranking did not change between the filter paper and artificial soil tests. The herbicide’s mixed effect in both test systems showed a consistent antagonistic effect relative to a Concentration Addition reference model. It indicates that the mixtures retracted the herbicide’s action in the earthworms.