Mycobacterium tuberculosis 18b, a streptomycin (STR)-dependent mutant that enters a viable but nonreplicating state in the absence of STR, has been developed as a simple model for drug testing against dormant bacilli. Here, we further evaluated the STR-starved 18b (SS18b) model both in vitro and in vivo by comparing the behavior of 22 approved and experimental tuberculosis drugs. Using the resazurin reduction microplate assay (REMA), rifampin (RIF), rifapentine (RPT), TMC207, clofazimine (CFM), and linezolid (LIN) were found to be active against SS18b in vitro, and their bactericidal activity was confirmed by determining the number of CFU. A latent 18b infection was established in mice, and some of the above-mentioned drugs were used for treatment, either alone or in combination with RIF. RIF, RPT, TMC207, CFM, and pyrazinamide (PZA) were all active in vivo, while cell wall inhibitors were not. A comparative kinetic study of rifamycin efficacy was then undertaken, and the results indicated that RPT clears latent 18b infection in mice faster than RIF. Intrigued by the opposing responses of live and dormant 18b cells to cell wall inhibitors, we conducted a systematic analysis of 14 such inhibitors using REMA. This uncovered an SS18b signature (CWPRED) that accurately predicted the activities of cell wall inhibitors and performed well in a blind study. CWPRED will be useful for establishing the mode of action of compounds with unknown targets, while the SS18b system should facilitate the discovery of drugs for treating latent tuberculosis.