<p>Industrial composition is one of the fundamental properties of biochar. In conventional near-infrared (NIR) spectroscopy technology for rapid, noninvasive analysis, the NIR spectral signal may vary with the variation of the spectrometer. Effective calibration transfer is essential for improving the applicability of NIR models in proximate analysis of biochar. In this study, 85 copyrolysis biochar samples (animal manures and crop residues, 300–700&#xa0;°C) were prepared. Optimized NIR PLS models with high accuracy (high correlation coefficient of prediction (<i>R</i>^<i>2</i>_<i>P</i>), low root mean square error of prediction (<i>RMSEP</i>), and high ratio of prediction to deviation (<i>RPD</i>)) based on three different NIR spectrometers were developed to quantify the contents of ash (&gt;0.86, &lt;2.31, and &gt;2.6), volatile matter (VM) (&gt;0.97, &lt;2.24, and &gt;5.0), and fixed carbon (FC) (&gt;0.87, &lt;2.55, and &gt;2.7) in biochar. Four calibration model transfer approaches with the spectrometers were tested using the chemometric algorithms of spectral subtraction correction (SSC), direct standardization (DS), piecewise direct standardization (PDS), and slope/bias correction (SBC). In terms of accuracy, the 3 spectrometers exhibited equivalent performance. The results were comparable to those obtained by the original NIR spectral models when the models were transferred across the three NIR spectrometers using a suitable method (for <i>RMSEP</i> value increases (%), ash: &lt;47.5%, VM: &lt;85.0%, and FC: &lt;-26.7%). Generally, the orders of accuracy of the optimized NIR PLS models and the transfer models were the same, i.e., VM &gt; FC and ash. The transfer performance of the models varied with different industrial compositions, transfer methods and NIR instruments.</p> Graphical abstract <p></p>