According to the fifth edition of the WHO Classification of Tumours of the Central Nervous System (CNS) published in 2021, grade 4 gliomas classification includes IDH-mutant astrocytomas and wild-type IDH glioblastomas. Unfortunately, despite precision oncology development, the prognosis for patients with grade 4 glioma remains poor, indicating an urgent need for better diagnostic and therapeutic strategies. Circulating miRNAs besides being important regulators of cancer development could serve as promising diagnostic biomarkers for patients with grade 4 glioma. Here, we propose a two-miRNA miR-362-3p and miR-6721-5p screening signature for serum for non-invasive classification of identified glioma cases into the highest-grade 4 and lower-grade gliomas. A total of 102 samples were included in this study, comprising 78 grade 4 glioma cases and 24 grade 2–3 glioma subjects. Using the NanoString platform, seven miRNAs were identified as differentially expressed (DE), which was subsequently confirmed via RT-qPCR analysis. Next, numerous combinations of DE miRNAs were employed to develop classification models. The dual panel of miR-362-3p and miR-6721-5p displayed the highest diagnostic value to differentiate grade 4 patients and lower grade cases with an AUC of 0.867. Additionally, this signature also had a high AUC = 0.854 in the verification cohorts by RT-qPCR and an AUC = 0.842 using external data from the GEO public database. The functional annotation analyses of predicted DE miRNA target genes showed their primary involvement in the STAT3 and HIF-1 signalling pathways and the signalling pathway of pluripotency of stem cells and glioblastoma-related pathways. For additional exploration of miRNA expression patterns correlated with glioma, we performed the Weighted Gene-Co Expression Network Analysis (WGCNA). We showed that the modules most associated with glioma grade contained as many as six DE miRNAs. In conclusion, this study presents the first evidence of serum miRNA expression profiling in adult-type diffuse glioma using a classification based on the WHO 2021 guidelines. We expect that the discovered dual miR-362-3p and miR-6721-5p signatures have the potential to be utilised for grading gliomas in clinical applications.