Responses to potassium (K) fertilizer are reported in different parts of Ethiopia in recent years, contrary to the general perception that K fertilizer was not necessary on the soils. However, use of adequate K level for maximum economic yield is possible only when soil test data are calibrated against response curve for a given crop and soil type. This pot trial was conducted to calibrate K soil test with the response of teff (Eragrostis tef (Zucc.)) at application of increasing levels of K. Random soil samples (0-20 cm) were collected from 60 locations having Mehlich-3 (M-3) K soil test values ranging from 16 to 910 mg kg −1 across 20 teff growing districts in the highlands of Ethiopia. The relative biomass yield of teff was significantly correlated with Mehlich-3 extractable K, suggesting that the extraction method fairly estimated the available soil K. Soil test M-3 extractable K values were categorized, based on relative biomass yield, as "very low" (< 55 mg kg −1), "low" (55-210 mg kg −1), "medium" (210-280 mg kg −1), "high" (280-500 mg kg −1) and "very high" (> 500 mg kg −1). And K application rates of 139 and 54 kg ha −1 for the very low and low K categories respectively would suffice for optimum teff yield. The critical limit of K in soil (M-3) using the graphical method was 210 mg kg −1 for the 85% relative biomass yield level. An inverse relation of yield response to the soil K status was observed suggesting that application of K fertilizer increases yield of teff in low status soils. However, field verification is required to draw sound conclusions.