BACKGROUND: Humanity has become largely dependent on artemisinin derivatives for both the treatment and control of malaria, with few alternatives available. A Plasmodium falciparum phenotype with delayed parasite clearance during artemisinin-based combination therapy has established in Southeast Asia, and is emerging elsewhere. Therefore, we must know how fast, and by how much, artemisinin-resistance can strengthen. METHODS: P. falciparum was subjected to discontinuous in vivo artemisinin drug pressure by capitalizing on a novel model that allows for long-lasting, high-parasite loads. Intravenous artesunate was administered, using either single flash-doses or a 2-day regimen, to P. falciparum-infected humanized NOD/SCID IL-2Rγ-/-immunocompromised mice, with progressive dose increments as parasites recovered. The parasite's response to artemisinins and other available anti-malarial compounds was characterized in vivo and in vitro. RESULTS: Artemisinin resistance evolved very rapidly up to extreme, near-lethal doses of artesunate (240 mg/kg), an increase of > 3000-fold in the effective in vivo dose, far above resistance levels reported from the field. Artemisinin resistance selection was reproducible, occurring in 80% and 41% of mice treated with flash-dose and 2-day regimens, respectively, and the resistance phenotype was stable. Measuring in vitro sensitivity proved inappropriate as an early marker of resistance, as IC50 remained stable despite in vivo resistance up to 30 mg/kg (ART-S: 10.7 nM (95% CI 10.2-11.2) vs. ART-R30: 11.5 nM (6.6-16.9), F = 0.525, p = 0.47). However, when in vivo resistance strengthened further, IC50 increased 10-fold (ART-R240 100.3 nM (92.9-118.4), F = 304.8, p < 0.0001), reaching a level much higher than ever seen in clinical samples. Artemisinin resistance in this African P. falciparum strain was not associated with mutations in kelch-13, casting doubt over the universality of this genetic marker for resistance screening. Remarkably, despite exclusive exposure to artesunate, full resistance to quinine, the only other drug sufficiently fast-acting to deal with severe malaria, evolved independently in two parasite lines exposed to different artesunate regimens in vivo, and was confirmed in vitro. CONCLUSION: P. falciparum has the potential to evolve extreme artemisinin resistance and more complex patterns of multidrug resistance than anticipated. If resistance in the field continues to advance along this trajectory, we will be left with a limited choice of suboptimal treatments for acute malaria, and no satisfactory option for severe malaria.