MODEL#llm#odia#apache-2.0
Odia-Llama-3 8B (instruct)
Fine-tuned on 1.4M Odia instruction pairs. Outperforms base Llama-3 on Odia benchmarks by 38%.
3,120 viewsPublished 5 Apr 2026
About this listing
Open-weights instruct model fine-tuned by the Odia GenAI working group. Apache 2.0. Includes eval card, training recipe, and Colab notebook for inference.