Abstract Purpose To investigate clinical usefulness of eribulin-based neoadjuvant chemotherapy in triple-negative breast cancer (TNBC) patients. Methods Patients in group A (aged < 65 years with homologous recombination deficiency, HRD, score ≥ 42, or those at any age with germline BRCA mutation, gBRCAm) were randomized to 4 cycles of paclitaxel plus carboplatin (group A1) or eribulin plus carboplatin (group A2), followed by 4 cycles of anthracycline. Patients in group B (aged < 65 years with HRD score < 42, or aged ≥ 65 years without gBRCAm) were randomized to 6 cycles of eribulin plus cyclophosphamide (group B1) or eribulin plus capecitabine (group B2); non-responders to the first 4 cycles of the eribulin-based therapy received anthracycline. Primary endpoint was pCR rate (ypT0-is, ypN0; centrally confirmed). Main secondary endpoint was safety. Results The full analysis set comprised 99 patients. The pCR rate was 65% (90% CI, 46%–81%) and 45% (27%–65%) in groups A1 and A2, respectively, and 19% (8%–35%) in both groups B1 and B2. No major difference was seen in secondary endpoints, but peripheral neuropathy incidence was 74% in group A1, whereas it was 32%, 22%, and 26% in groups A2, B1, and B2, respectively. Conclusions In patients aged < 65 years with high HRD score or gBRCAm, weekly paclitaxel plus carboplatin and eribulin plus carboplatin followed by anthracycline resulted in a pCR rate of > 60% and > 40%, respectively, suggesting potential usefulness of patient stratification using HRD; pCR tended to be low in patients with HRD-negative tumors. Neurotoxicity was less frequent with the eribulin-based regimen. Trial registration:The study has been registered with the University Hospital Medical Information Network Clinical Trials Registry (http://www.umin.ac.jp/ctr/index-j.htm) with unique trial number UMIN000023162. The Japan Breast Cancer Research Group trial number is JBCRG-22.