Combination therapies containing Artemisinin derivatives (ART) are the first line treatment of malaria but their effectiveness is reduced by resistance due to a fraction of ring stage Plasmodium parasites surviving high ART levels. Resistance-causing kelch13 (k13) mutations are common in South East Asia but were only recently and sparsely detected in Africa, which experiences the highest malaria burden. Kelch13 shares a cellular compartment with other proteins (KICs) several of which cause resistance when inactivated or, in a few cases, when mutated. To see if the sparse detection of resistance mutations in Africa is due to unknown mutations, we tested 135 k13 and kic mutations detected mostly in African field isolates. No kic mutation caused resistance but two in k13, V520A and V589I, did. These mutations were geographically much more widespread in Africa but conferred lower levels of resistance than known ART resistance mutations. A dissection of the mechanism using isogenic parasites with different k13 mutations and parasites that we selected for even higher ART resistance, showed that resistance is a function of K13 protein levels in ring-stage parasites and correlates with the fitness cost. This indicated that hyper-resistance is unlikely to arise in the field. Double mutations in k13 had not even additive effects and combinations including a non-k13 mutation led to high fitness costs, suggesting that such combinations also pose no risk for higher resistance in the field. Overall, our results indicate that resistance is restricted by a proportional fitness cost but that incidence-lowering measures may favor high-resistance mutations.SignificanceOur findings indicate that hyper-resistant parasites are unlikely to occur in endemic settings due to the proportional fitness cost. Mutations within k13 were not additive and mutations outside k13 had a disproportionally high fitness cost. The strong influence of the fitness cost may have favored moderate frequencies of the resistance mutations with low fitness cost detected in many African settings that we characterized in this study. These mutations, so far gone unnoticed, may be optimal in high endemicity regions, where relative drug use is presumably low but frequent multiple infections increase competition. A given endemic setting may thus favor variants with K13 levels for an optimal resistance-fitness balance which is relevant for incidence-lowering interventions.