Hybrid Mamba2–Transformer MoE, 30B total / ~3.5B active, trained end-to-end for Indic capability across 11 languages. 128K context, tool calling, native-script output. Apache-2.0.
gnani-evon-v3-30B-A3B is a hybrid Mamba–Transformer Mixture-of-Experts model from Gnani.ai, trained end-to-end for Indic capability. 30B total parameters with ~3B active per token, 128K context, tool calling, and native-script support across English, Hindi, Bengali, Telugu, Tamil, Marathi, Gujarati, Kannada, Malayalam, Odia, and Punjabi. Training ran on a blended English + Indic corpus through continued pretraining, supervised fine-tuning, and reinforcement learning (GRPO), with every stage oriented toward native-script comprehension, extractive QA, reasoning, and instruction following. Intended use: multilingual chat, RAG, assistants, and agentic workflows in Indic languages and English.
Apache 2.0
Gnani AI Research Team
Text Generation
PyTorch
Open
Sector Agnostic
16/08/26 04:47:34
0
Apache 2.0
© 2026 - Copyright AIKosh. All rights reserved.