T cell-mediated immunity is crucial for protection against severe COVID-19. The performance of SARS-CoV-2-specific IFN-γ release assays (IGRAs) to measure T cell responses during severe acute infection is unknown. We conducted a prospective, longitudinal analysis of patients hospitalized for confirmed COVID-19. A standardized SARS-CoV-2-specific quantitative IGRA was measured on admission. Follow-up lasted 90 days. Two hundred forty-eight patients were included; 181 (73%) were vaccinated, 142 (57.3%) were infected with the Omicron variant, and 106 (42.7%) were infected with the Delta variant. The SARS-CoV-2-IGRA result was positive (>200 mIU/mL) in 125 (50.5%) patients, with 44 (35.2%) showing concomitant mitogen-induced IFN-γ response. Compared with patients with negative and borderline results, those with positive SARS-CoV-2 IGRA were younger and showed lower frequency of co-existing comorbidities, lower median Charlson comorbidity index, lower frequency of World Health Organization (WHO) severity score > 4, higher peripheral arterial oxygen saturation to inspired fraction of oxygen (SpO2/FiO2) ratio on admission, higher S-IgG and N-IgG antibody responses, and lower mortality at 28/60/90 days. These findings were consistent for both the Omicron and Delta variants. Similar patterns to those described above were observed in the subset of patients with positive SARS-CoV-2-IGRA and indeterminate mitogen-induced responses, as well as in those exhibiting both positive SARS-CoV-2-specific and mitogen responses. Twenty-eight (11.3%) patients had borderline (100-200 mIU/mL) SARS-CoV-2-IGRA results and shared several similarities with patients with negative SARS-CoV-2 IGRA. In an adjusted Cox model, a negative SARS-CoV-2-IGRA result was found to be an independent predictor of mortality. Using a receiver operating characteristics curve analysis, we found that an IFN-γ value of 150 mIU/mL was identified as the optimal cut-off point for predicting mortality, with 79% [95% confidence interval (CI) 61-93] sensitivity, 61% (95% CI 55-68) specificity, 18% (95% CI 15-22) positive predictive value, 96% (95% CI 93-98) negative predictive value, and an area under the curve (AUC) of 70%. The assay's performance remained consistent regardless of mitogen results or the presence of the Delta or Omicron variants. IMPORTANCE The cellular immune response is essential in the protection against severe disease in patients with established SARS-CoV-2 infection. The novelty of this study lies in the evaluation of the overall performance of a standardized assay to measure cellular immune response, the SARS-CoV-2-specific interferon-γ release assay (IGRA), in hospitalized patients with severe COVID-19. The SARS-CoV-2 IGRA was shown to accurately classify patients based on disease severity and prognosis, and the study revealed that test performance was not affected by the SARS-CoV-2 variant or control tube results. We identified an assay cut-off point with a high negative predictive value against mortality. The SARS-CoV-2 IGRA in patients hospitalized for COVID-19 may be a useful tool to assess cellular immunity and adopt targeted therapeutic and preventive measures.