Cyanide compounds are extensively used in industries like mining, metallurgy, and chemical synthesis, but their high toxicity presents serious environmental and health risks. This study applies advanced modeling techniques such as Quantitative Structure-Toxicity Relationship (QSTR), Species cyanide-Sensitivity Distribution (ScSD), and quantitative Read-Across Structure Toxicity (q-RASTR) to assess cyanide toxicity. A dataset of 25 cyanide salts was analyzed for acute, chronic, and lethal toxicity across species like humans, rats, and fish. Key molecular descriptors, including topological, geometrical, and electronic properties, were computed using ALOGPS 2.1, ChemAxon, and Elemental-Descriptor 1.0. Three machine learning methods MLR, PLS, and kNN were employed to develop predictive models. Further, q-RASTR models were developed to enhance the predictive power by similarity measures concept of the studied cyanides by integrating features from QSTR and ScSD models. These models were validated using external datasets, achieving high accuracy. Key descriptors such as refractivity, water solubility, and lipophilic components significantly influence cyanide toxicity. The combined QSTR, ScSD, and q-RASTR models provide a robust framework for predicting species-specific cyanide-sensitivity, enhancing our understanding of cyanide's molecular toxicity mechanisms. This research aids environmental risk assessment and informs safer regulatory strategies. The results are available for public access at https://nanosens.onrender.com/apps/calTox/index.html#/.