Instructions to use md-nishat-008/TigerLLM-1B-it with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use md-nishat-008/TigerLLM-1B-it with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="md-nishat-008/TigerLLM-1B-it")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("md-nishat-008/TigerLLM-1B-it")
model = AutoModelForCausalLM.from_pretrained("md-nishat-008/TigerLLM-1B-it", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use md-nishat-008/TigerLLM-1B-it with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "md-nishat-008/TigerLLM-1B-it"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "md-nishat-008/TigerLLM-1B-it",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/md-nishat-008/TigerLLM-1B-it

SGLang

How to use md-nishat-008/TigerLLM-1B-it with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "md-nishat-008/TigerLLM-1B-it" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "md-nishat-008/TigerLLM-1B-it",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "md-nishat-008/TigerLLM-1B-it" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "md-nishat-008/TigerLLM-1B-it",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use md-nishat-008/TigerLLM-1B-it with Docker Model Runner:
```
docker model run hf.co/md-nishat-008/TigerLLM-1B-it
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

🐯 TigerLLM-1B-it

A compact Bangla instruction-tuned LLM from the TigerLLM family (ACL 2025)

Nishat Raihan, Marcos Zampieri · George Mason University

The LLaMA-based TigerLLM 1B is deprecated. The 1B model in the ACL 2025 paper was built on LLaMA 3.2. Those checkpoints are obsolete and no longer maintained. This repository hosts the Gemma 3 version of TigerLLM-1B-it, which scores 2–4% higher than the LLaMA-based 1B on the benchmarks below.

TigerLLM-9B-it is deprecated too. For the larger model, use TigerLLM-12B-it, which scores 5–10% higher than the 9B model.

Model summary

TigerLLM-1B-it is a small Bangla instruction-following model. It answers questions, explains, rewrites, summarizes and translates in Bangla, and it is light enough for a laptop GPU or a CPU. It introduces itself as TigerLLM and knows who built it.


Architecture	`Gemma3ForCausalLM` (Gemma 3, 1B)
Parameters	1.0B
Precision	BF16, one `model.safetensors` file (about 1.9 GiB)
Context length	32,768 tokens (from `config.json`)
Languages	Bangla (primary), English
Fine-tuning data	Bangla-Instruct
License	CC BY 4.0

Benchmarks

Scores on six Bangla benchmarks, from Table 2 of the TigerLLM paper (ACL 2025).

Model	MMLU-bn	PangBench-bn	BanglaQuaD	mHumanEval-bn	BEnQA	BanglaRQA
GPT-3.5	0.55	0.55	0.50	0.56	0.50	0.49
Gemini 1.5 Flash	0.66	0.57	0.62	0.58	0.56	0.61
GPT-4o-mini	0.67	0.62	0.65	0.56	0.60	0.60
LLaMA 3.2 (11B)	0.22	0.19	0.21	0.15	0.18	0.20
Gemma 2 (27B)	0.35	0.51	0.43	0.64	0.50	0.56
Pangea (7B)	0.18	0.15	0.17	0.10	0.14	0.16
Titu-LLM	0.06	0.19	0.08	0.02	0.17	0.21
Bong-LLaMA	0.05	0.12	0.08	0.02	0.15	0.13
Bangla-LLaMA	0.02	0.08	0.05	0.10	0.11	0.09
Bangla-Gemma	0.18	0.15	0.12	0.10	0.22	0.19
TigerLLM 1B, LLaMA 3.2 (deprecated)	0.61	0.55	0.68	0.61	0.59	0.62
TigerLLM-1B-it, Gemma 3 (this model)	+2–4%	+2–4%	+2–4%	+2–4%	+2–4%	+2–4%
TigerLLM-9B-it (deprecated)	0.72	0.68	0.70	0.63	0.65	0.68
TigerLLM-12B-it	+5–10%	+5–10%	+5–10%	+5–10%	+5–10%	+5–10%

The numeric rows are from the paper. This model scores 2–4% higher than the LLaMA-based TigerLLM 1B, and TigerLLM-12B-it scores 5–10% higher than TigerLLM-9B-it.

Quick start

pip install "torch>=2.4" "transformers>=4.56" accelerate sentencepiece

The model needs about 2 GB of GPU memory in BF16, and it also runs on CPU or Apple Silicon (device_map="mps").

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

MODEL_ID = "md-nishat-008/TigerLLM-1B-it"

tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID,
    dtype=torch.bfloat16,
    device_map="auto",
).eval()

messages = [{"role": "user", "content": "বাংলায় সহজ ভাষায় ব্যাখ্যা করুন: মেশিন লার্নিং কী এবং এটি কোথায় ব্যবহৃত হয়?"}]

inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_dict=True,
    return_tensors="pt",
).to(model.device)

with torch.inference_mode():
    output_ids = model.generate(**inputs, max_new_tokens=256, do_sample=False, repetition_penalty=1.05)

print(tokenizer.decode(output_ids[0, inputs["input_ids"].shape[-1]:], skip_special_tokens=True))

Or with a pipeline:

from transformers import pipeline

chat = pipeline("text-generation", model="md-nishat-008/TigerLLM-1B-it", dtype="bfloat16")
print(chat([{"role": "user", "content": "তোমার নাম কী?"}], max_new_tokens=128)[0]["generated_text"][-1]["content"])

Prompting tips

Build prompts with tokenizer.apply_chat_template(...). Do not paste special chat tokens by hand.
Say the language, task, audience and output format in the prompt, for example: বাংলায় ৮ম শ্রেণির শিক্ষার্থীর উপযোগী করে সালোকসংশ্লেষণ ব্যাখ্যা করুন। উত্তরটি তিনটি ছোট অনুচ্ছেদে দিন।
Use do_sample=False for reproducible output. The default generation config samples with top_k=64 and top_p=0.95.
For multi-turn chat, alternate user and assistant messages and end with the new user turn.

Training

The current checkpoint was trained in two stages on top of Gemma 3 (1B): instruction tuning on Bangla-Instruct, then a short identity fine-tune.

Identity fine-tune	Value
Data	2,000 TigerLLM identity examples (Bangla and English) mixed with 6,000 Bangla-Instruct examples, so earlier skills are kept
Method	LoRA (rank 16, alpha 32) on all attention and MLP projections, merged into the weights
Epochs / learning rate	2 / 1e-4, cosine schedule
Check	loss on 500 held-out Bangla-Instruct examples went from 1.383 to 1.356

Fine-tuning setup the paper reports for the LLaMA-based 1B model

Parameter	Value
Max sequence length	2048
Batch size (train / eval)	16
Gradient accumulation steps	4
Epochs	3
Learning rate	1e-5
Weight decay	0.02
Warm-up	10% of steps
Optimizer	AdamW (8-bit)
LR scheduler	Cosine
Precision	BF16
Seed	42

Training data

Both datasets come from the TigerLLM paper.

Bangla-Instruct: native Bangla instruction-response pairs. 50 volunteers from Bangladeshi universities wrote 500 seed tasks across 10 categories (culture, academic writing, math, programming, creative writing, science, business, social issues, data analysis, translation). GPT-4 and Claude 3.5 Sonnet expanded them with self-instruct, and every pair passed filters for language, cultural sensitivity, content quality and novelty.
Bangla-TextBook: 9.9M tokens from 163 open textbooks (Grades 6–12) published by the National Curriculum and Textbook Board of Bangladesh. The paper's models used it for continual pretraining.

Repository files

File	Contents
`model.safetensors`	Full BF16 model weights
`config.json`, `generation_config.json`	Model and generation settings
`tokenizer.json`, `tokenizer.model`, `tokenizer_config.json`, `added_tokens.json`, `special_tokens_map.json`, `chat_template.jinja`	Tokenizer and chat format

Limitations

The model can hallucinate and can produce outdated, biased, unsafe or wrong answers. A 1B model makes more of these mistakes than TigerLLM-12B-it, especially on arithmetic and multi-step reasoning.
Quality varies across dialects, romanized Bangla, specialized domains and long inputs.
Do not use it as the only source for medical, legal, financial or other high-stakes decisions. Keep a human in the loop.

License

CC BY 4.0. Check the Gemma 3 terms of use and the dataset licenses before redistribution or commercial use.

Citation

@inproceedings{raihan-zampieri-2025-tigerllm,
    title = "{T}iger{LLM} - A Family of {B}angla Large Language Models",
    author = "Raihan, Nishat  and
      Zampieri, Marcos",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 2: Short Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.acl-short.69/",
    doi = "10.18653/v1/2025.acl-short.69",
    pages = "887--896",
    ISBN = "979-8-89176-252-7"
}

Questions: md.nishat.008@gmail.com

Downloads last month: 1,897

Safetensors

Model size

1.0B params

Tensor type

BF16

Model tree for md-nishat-008/TigerLLM-1B-it

Adapters

2 models

Finetunes