Instructions to use GAIR/LIMI with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use GAIR/LIMI with Transformers:

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

pipe = pipeline("text-generation", model="GAIR/LIMI")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("GAIR/LIMI")
model = AutoModelForCausalLM.from_pretrained("GAIR/LIMI", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use GAIR/LIMI with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/GAIR/LIMI

SGLang

How to use GAIR/LIMI 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 "GAIR/LIMI" \
    --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": "GAIR/LIMI",
		"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 "GAIR/LIMI" \
        --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": "GAIR/LIMI",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use GAIR/LIMI with Docker Model Runner:
```
docker model run hf.co/GAIR/LIMI
```

mhjiang0408 commited on Oct 9, 2025

Commit

5d648f0

verified ·

1 Parent(s): 39957da

Update README.md

Browse files

Files changed (1) hide show

README.md +15 -1

README.md CHANGED Viewed

@@ -52,7 +52,21 @@ LIMI is an agentic model fine‑tuned from [GLM‑4.5](https://huggingface.co/za
 - Training framework: slime
 - Training data: curated conversations from [GAIR/LIMI](https://huggingface.co/datasets/GAIR/LIMI)
-## Performance on AgencyBench
 Our models achieve state-of-the-art performance across multiple agentic evaluation tasks:

 - Training framework: slime
 - Training data: curated conversations from [GAIR/LIMI](https://huggingface.co/datasets/GAIR/LIMI)
+## Performance
+### SFT with LIMI Dataset on Dense Models
+Our LIMI dataset significantly enhances dense models (Qwen3 series) on both in-domain and out-of-domain benchmarks:
+<p align="center">
+  <img src="./assets/generalize_improvement.png" style="width: 85%;" alt="Performance Improvements on AgencyBench and Out-of-Domain Benchmarks">
+</p>
+The figure above demonstrates the effectiveness of our training approach:
+- **Left (AgencyBench)**: Substantial improvements on in-domain agentic tasks, with Qwen3-4B (4.6% → 8.6%), Qwen3-8B (7.3% → 10.6%), and Qwen3-32B (8.4% → 20.5%).
+- **Right (Out-of-Domain)**: Strong generalization to unseen benchmarks while maintaining performance, with Qwen3-4B (28.3% → 28.9%), Qwen3-8B (31.2% → 32.0%), and Qwen3-32B (35.2% → 37.1%).
+### LIMI Models on AgencyBench
 Our models achieve state-of-the-art performance across multiple agentic evaluation tasks: