Instructions to use meituan-longcat/LongCat-Flash-Thinking with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use meituan-longcat/LongCat-Flash-Thinking with Transformers:

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

pipe = pipeline("text-generation", model="meituan-longcat/LongCat-Flash-Thinking", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("meituan-longcat/LongCat-Flash-Thinking", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use meituan-longcat/LongCat-Flash-Thinking with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/meituan-longcat/LongCat-Flash-Thinking

SGLang

How to use meituan-longcat/LongCat-Flash-Thinking 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 "meituan-longcat/LongCat-Flash-Thinking" \
    --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": "meituan-longcat/LongCat-Flash-Thinking",
		"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 "meituan-longcat/LongCat-Flash-Thinking" \
        --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": "meituan-longcat/LongCat-Flash-Thinking",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use meituan-longcat/LongCat-Flash-Thinking with Docker Model Runner:
```
docker model run hf.co/meituan-longcat/LongCat-Flash-Thinking
```

LongCat0830

wjn1996 commited on Sep 24, 2025

Commit

7c9aad1

verified ·

1 Parent(s): 55f5777

Update configuration_longcat_flash.py (#5)

Browse files

- Update configuration_longcat_flash.py (4357d408aa0bfbac7d38b5cea2bda88745a338b5)

Co-authored-by: WangJianing <wjn1996@users.noreply.huggingface.co>

Files changed (1) hide show

configuration_longcat_flash.py +2 -3

configuration_longcat_flash.py CHANGED Viewed

@@ -1,4 +1,3 @@
 """LongcatFlash model configuration"""
 from transformers.configuration_utils import PretrainedConfig
@@ -53,7 +52,7 @@ class LongcatFlashConfig(PretrainedConfig):
             Dimension of the value heads.
         qk_nope_head_dim (`int`, *optional*, defaults to 128):
             Dimension of the query/key heads that don't use rotary position embeddings.
-        norm_topk_prob (`bool`, *optional*, defaults to `True`):
             Whether to normalize the weights of the routed experts.
         hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
             The non-linear activation function (function or string) in the decoder.
@@ -213,4 +212,4 @@ class LongcatFlashConfig(PretrainedConfig):
         return self.num_layers
-__all__ = ["LongcatFlashConfig"]

 """LongcatFlash model configuration"""
 from transformers.configuration_utils import PretrainedConfig
             Dimension of the value heads.
         qk_nope_head_dim (`int`, *optional*, defaults to 128):
             Dimension of the query/key heads that don't use rotary position embeddings.
+        norm_topk_prob (`bool`, *optional*, defaults to `False`):
             Whether to normalize the weights of the routed experts.
         hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
             The non-linear activation function (function or string) in the decoder.
         return self.num_layers
+__all__ = ["LongcatFlashConfig"]