The 29M-parameter LoRA adapter (r=16) behind pankajpandey-dev/qwen3.5-9b-hindi-instruct. Attach to base Qwen3.5-9B, or use as a starting point for your own Indic fine-tunes.

python

from peft import PeftModel
from transformers import AutoModelForCausalLM
base = AutoModelForCausalLM.from_pretrained("unsloth/Qwen3.5-9B", torch_dtype="bfloat16", device_map="auto")
model = PeftModel.from_pretrained(base, "pankajpandey-dev/qwen3.5-9b-hindi-instruct-lora")

Full recipe, data mix and limitations: see the main model card.

🇮🇳 About the Hindi LLM Series

Weekly open releases making small LLMs speak fluent, native Hindi — trained on free/low-cost GPUs, shipped as GGUF for laptops and edge devices. Built by pankajpandey-dev (contact links on profile).

This release: Model · GGUF · LoRA · Series: 🇮🇳 Hindi LLM Collection