Timber is becoming a popular construction material even for high-rise buildings despite its poorly understood fire behaviour. In a fire, timber—a natural polymer—degrades in the thermochemical process of charring, causing it to lose structural strength. In spite of significant research on the physics of charring, the chemical kinetics—reactions and kinetic parameters for pyrolysis and oxidation—remains a scientific challenge to model accurately. Current kinetic models are either computationally too expensive or neglect key chemical pathways. Here we derive a new appropriate kinetic model for fire science at the microscale using a novel methodology. First, we built a kinetic model for each component of timber (cellulose, hemicellulose, and lignin) from literature studies and experiments of the components. Then, we combined these three models into one kinetic model (8 reactions, 8 chemical species) for timber. This approach accounts for chemical differences among timber species. However, the timber model is only able to reproduce the trend in the experiments when literature parameters are used. Using multi-objective inverse modelling, we extract a new set of optimised kinetic parameters from 16 high-quality experiments from the literature. The novel optimised kinetic model is able to reproduce these 16 and a further 64 (blind predictions) experiments nearly within the experimental uncertainty, spanning different heating rates (1–60 K/min), oxygen concentrations (0–60 %), and even isothermal experiments (220–300 °C). Furthermore, the model outperforms current kinetic models for fire science in accuracy across a wide range of conditions without an increase in complexity. Incorporated into a model of heat and mass transfer, this new and optmised kinetic model could improve the understanding of timber burning and has the potenial to lead to safer designs of timber buildings.