A new approach based on pulsed photoacoustic spectroscopy for non-invasively quantifying tissue chromophore concentrations with high spatial resolution has been developed. The technique is applicable to the quantification of tissue chromophores such as oxyhaemoglobin (HbO₂) and deoxyhaemoglobin (HHb) for the measurement of physiological parameters such as blood oxygen saturation (SO₂) and total haemoglobin concentration. It can also be used to quantify the local accumulation of targeted contrast agents used in photoacoustic molecular imaging. The technique employs a model-based inversion scheme to recover the chromophore concentrations from photoacoustic measurements. This comprises a numerical forward model of the detected time-dependent photoacoustic signal that incorporates a multiwavelength diffusion-based finite element light propagation model to describe the light transport and a time-domain acoustic model to describe the generation, propagation and detection of the photoacoustic wave. The forward model is then inverted by iteratively fitting it to measurements of photoacoustic signals acquired at different wavelengths to recover the chromophore concentrations. To validate this approach, photoacoustic signals were generated in a tissue phantom using nanosecond laser pulses between 740 nm and 1040 nm. The tissue phantom comprised a suspension of intralipid, blood and a near-infrared dye in which three tubes were immersed. Blood at physiological haemoglobin concentrations and oxygen saturation levels ranging from 2% to 100% was circulated through the tubes. The signal amplitude from different temporal sections of the detected photoacoustic waveforms was plotted as a function of wavelength and the forward model fitted to these data to recover the concentrations of HbO₂ and HHb, total haemoglobin concentration and SO₂. The performance was found to compare favourably to that of a laboratory CO-oximeter with measurement resolutions of ±3.8 g l⁻¹ (±58 µM) and ±4.4 g l⁻¹ (±68 µM) for the HbO₂ and HHb concentrations respectively and ±4% for SO₂ with an accuracy in the latter in the range −6%–+7%.