Introduction – Diffuse gliomas comprise a heterogeneous group of primary brain tumors with different biological behavior and aggressiveness. Intraoperative knowledge of tumor subtype and grade could enable adaptation of the surgical strategy. In this context, label-free multiphoton imaging modalities that allow in situ tissue characterization for glioma grading were investigated. Methods – A multimodal imaging approach combining coherent anti-Stokes Raman scattering (CARS), autofluorescence (AF), and second harmonic generation (SHG) was applied to acquire an extensive dataset from 102 glioma patients, including glioblastoma (WHO grade 4), astrocytoma (WHO grades 2, 3, and 4), and oligodendroglioma (WHO grades 2 and 3). The prevalence of specific tissue features was evaluated based on visual inspection and used to build a score for tumor malignancy. Results – CARS detected lipid droplets, AF showed cellular structures, and SHG-visualized blood vessels as well as remodeling of extracellular matrix. These were identified as histological features associated with tumor malignancy and recognizable intraoperatively by the surgeon. Quantification showed statistically significant differences in feature prevalence among tumor types, particularly between WHO grade 2 and grade 4 gliomas, despite substantial variability across patients and within individual tissue samples. Integration of multiple imaging features into a numerical score based on their presence or absence yielded values indicative of malignancy, with the highest scores observed exclusively in WHO grade 4 tumors. Conclusion – Label-free multiphoton imaging shows strong potential as an intraoperative diagnostic tool, enabling surgeons to score readily identifiable tissue features and obtain enhanced real-time information on glioma malignancy.