ResearchBenchmarksTokenization Efficiency on Urdu (urd)Follow1Bytes per TokenSarvam-2B0.8182.04653.2754.5035Nov 5, 2025Evaluation ResultsMethodMethodLinksBytes per TokenSarvam-2B2025.111LLaMA-42025.114.54Sutra2025.115.15GPT-oss2025.115.28Gemma-32025.115.55MUTANT-Indic2025.115.55