Linezolid has an excellent sterilizing effect in tuberculosis patients but high adverse event rates. The dose that would maximize efficacy and minimize toxicity is unknown. We performed linezolid dose-effect and dose-scheduling studies in the hollow fiber system model of tuberculosis (HFS-TB) for sterilizing effect. HFS-TB units were treated with several doses to mimic human-like linezolid intrapulmonary pharmacokinetics and repetitively sampled for drug concentration, total bacterial burden, linezolid-resistant subpopulations, and RNA sequencing over 2 months. Linezolid-resistant isolates underwent whole-genome sequencing. The expression of genes encoding efflux pumps in the first 1 to 2 weeks revealed the same exposure-response patterns as the linezolid-resistant subpopulation. Linezolid-resistant isolates from the 2nd month of therapy revealed mutations in several efflux pump/transporter genes and a LuxR-family transcriptional regulator. Linezolid sterilizing effect was linked to the ratio of unbound 0- to 24-h area under the concentration-time curve (AUC0–24) to MIC. Optimal microbial kill was achieved at an AUC0–24/MIC ratio of 119. The optimal sterilizing effect dose for clinical use was identified using Monte Carlo simulations. Clinical doses of 300 and 600 mg/day (or double the dose every other day) achieved this target in 87% and >99% of 10,000 patients, respectively. The susceptibility breakpoint identified was 2 mg/liter. The simulations identified that a 300-mg/day dose did not achieve AUC0–24s associated with linezolid toxicity, while 600 mg/day achieved those AUC0–24s in <20% of subjects. The linezolid dose of 300 mg/day performed well and should be compared to 600 mg/day or 1,200 mg every other day in clinical trials.