Instructions to use moonshotai/Kimi-VL-A3B-Thinking-2506 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use moonshotai/Kimi-VL-A3B-Thinking-2506 with Transformers:

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

pipe = pipeline("image-text-to-text", model="moonshotai/Kimi-VL-A3B-Thinking-2506", trust_remote_code=True)
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("moonshotai/Kimi-VL-A3B-Thinking-2506", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use moonshotai/Kimi-VL-A3B-Thinking-2506 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "moonshotai/Kimi-VL-A3B-Thinking-2506"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "moonshotai/Kimi-VL-A3B-Thinking-2506",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/moonshotai/Kimi-VL-A3B-Thinking-2506

SGLang

How to use moonshotai/Kimi-VL-A3B-Thinking-2506 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 "moonshotai/Kimi-VL-A3B-Thinking-2506" \
    --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": "moonshotai/Kimi-VL-A3B-Thinking-2506",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "moonshotai/Kimi-VL-A3B-Thinking-2506" \
        --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": "moonshotai/Kimi-VL-A3B-Thinking-2506",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use moonshotai/Kimi-VL-A3B-Thinking-2506 with Docker Model Runner:
```
docker model run hf.co/moonshotai/Kimi-VL-A3B-Thinking-2506
```

issue with vLLM inference

by rohitg - opened Jul 14, 2025

Discussion

rohitg

Jul 14, 2025

I keep running into this issue when trying to do inference using vLLM with pytorch 2.7 cu118

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/vllm/compilation/cuda_piecewise_backend.py", line 111, in call

ERROR 07-14 16:50:13 [core.py:515] return self.compiled_graph_for_general_shape(*args)

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py", line 838, in _fn

ERROR 07-14 16:50:13 [core.py:515] return fn(*args, **kwargs)

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/torch/_functorch/aot_autograd.py", line 1201, in forward

ERROR 07-14 16:50:13 [core.py:515] return compiled_fn(full_args)

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/torch/_functorch/_aot_autograd/runtime_wrappers.py", line 328, in runtime_wrapper

ERROR 07-14 16:50:13 [core.py:515] all_outs = call_func_at_runtime_with_args(

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] out = normalize_as_list(f(args))

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] outs = compiled_fn(args)

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/torch/_functorch/_aot_autograd/runtime_wrappers.py", line 495, in wrapper

ERROR 07-14 16:50:13 [core.py:515] return compiled_fn(runtime_args)

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/torch/_inductor/output_code.py", line 460, in call

ERROR 07-14 16:50:13 [core.py:515] return self.current_callable(inputs)

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/torch/_inductor/utils.py", line 2404, in run

ERROR 07-14 16:50:13 [core.py:515] return model(new_inputs)

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] File "/home/ro308090/.cache/vllm/torch_compile_cache/1316656fd2/rank_0_0/inductor_cache/xw/cxww7bk4vsxx3nbg5o7zwbfp3lgkbpglqtnlg47ymw7kxtw3cklm.py", line 890, in call

ERROR 07-14 16:50:13 [core.py:515] torch.ops.vllm.inplace_fused_experts.default(reinterpret_tensor(buf17, (s0, 2048), (2048, 1), 0), arg9_1, arg10_1, buf18, buf19, 'silu', False, False, False, False, False, False, 64)

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/torch/_ops.py", line 756, in call

ERROR 07-14 16:50:13 [core.py:515] return self._op(*args, **kwargs)

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/vllm/model_executor/layers/fused_moe/fused_moe.py", line 1018, in inplace_fused_experts

ERROR 07-14 16:50:13 [core.py:515] fused_experts_impl(hidden_states, w1, w2, topk_weights, topk_ids, True,

ERROR 07-14 16:50:13 [core.py:515] moe_align_block_size(curr_topk_ids, config['BLOCK_SIZE_M'],

ERROR 07-14 16:50:13 [core.py:515] ops.moe_align_block_size(topk_ids, num_experts, block_size, sorted_ids,

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/vllm/_custom_ops.py", line 1519, in moe_align_block_size

ERROR 07-14 16:50:13 [core.py:515] torch.ops._moe_C.moe_align_block_size(topk_ids, num_experts, block_size,

ERROR 07-14 16:50:13 [core.py:515] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

ERROR 07-14 16:50:13 [core.py:515] File "/lustre/fs1/groups/mshah/reasoning_rohit/reasoning/tlqa/lib/python3.12/site-packages/torch/_ops.py", line 1267, in getattr

ERROR 07-14 16:50:13 [core.py:515] raise AttributeError(

ERROR 07-14 16:50:13 [core.py:515] AttributeError: '_OpNamespace' '_moe_C' object has no attribute 'moe_align_block_size'

youkaichao

Jul 15, 2025

can you give a full repro and report your environment? you can post a bug report issue in vllm.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment