Abstract Objective: This study aimed at comparing different recovery-based methods to assess the highest exercise oxygen uptake value ( V ˙ O₂peak) when swimming at low-moderate, heavy and severe intensities. Complementarily, the different recovery curve kinetics were analysed. Approach: Eighteen competitive swimmers performed a 5 × 200 m front crawl intermittent protocol (0.05 m · s⁻¹ increments and 3 min intervals), with respiratory gas exchange being continuously measured breath-by-breath during and post-exercise using a portable gas analyser. The directly determined V ˙ O₂peak ( V ˙ O_2dir) was compared with the values obtained by linear and exponential backward extrapolations (of different intervals) and the recovery curve mathematical modelling. Main results: V ˙ O_2dir rose with intensity increase: 41.96 ± 6.22, 46.36 ± 6.89 and 50.97 ± 7.28 ml · kg⁻¹ min⁻¹ for low-moderate, heavy and severe swims. Linear and exponential regressions applied to the first 20 s of recovery presented the V ˙ O₂peak values closest to V ˙ O_2dir at low-moderate (42.80 ± 5.54 vs 42.88 ± 5.58 ml kg⁻¹ min⁻¹), heavy (47.12 ± 4.91 vs 47.48 ± 5.09 ml kg⁻¹ min⁻¹) and severe intensity domains (51.24 ± 6.89 vs 53.60 ± 8.54 ml kg⁻¹ · min⁻¹, respectively; r = 0.5–0.8, p < 0.05). The mono-exponential function was the best fit at low-moderate and heavy intensities, while the bi-exponential function better characterized the severe exercise domain (with a slow component amplitude, time delay and time constant of 6.2 ± 2.3 ml kg⁻¹ min⁻¹, 116.6 ± 24.3 and 39.9 ± 15.2 s, respectively). Significance: The backward extrapolation of the first 20 s of recovery is the best method to assess the V ˙ O₂peak for a large spectrum of swimming intensities. Complementarily, intensity increases imply different recovery curve kinetics, particularly a mono-exponential behaviour for low-moderate and heavy exertions and a bi-exponential dynamics for severe paces.