Instructions to use MiniMaxAI/MiniMax-M3.1-Flash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MiniMaxAI/MiniMax-M3.1-Flash with Transformers:

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

pipe = pipeline("image-text-to-text", model="MiniMaxAI/MiniMax-M3.1-Flash", 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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("MiniMaxAI/MiniMax-M3.1-Flash", trust_remote_code=True)
model = AutoModelForMultimodalLM.from_pretrained("MiniMaxAI/MiniMax-M3.1-Flash", trust_remote_code=True, device_map="auto")
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?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use MiniMaxAI/MiniMax-M3.1-Flash with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MiniMaxAI/MiniMax-M3.1-Flash"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MiniMaxAI/MiniMax-M3.1-Flash",
		"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/MiniMaxAI/MiniMax-M3.1-Flash

SGLang

How to use MiniMaxAI/MiniMax-M3.1-Flash 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 "MiniMaxAI/MiniMax-M3.1-Flash" \
    --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": "MiniMaxAI/MiniMax-M3.1-Flash",
		"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 "MiniMaxAI/MiniMax-M3.1-Flash" \
        --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": "MiniMaxAI/MiniMax-M3.1-Flash",
		"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 MiniMaxAI/MiniMax-M3.1-Flash with Docker Model Runner:
```
docker model run hf.co/MiniMaxAI/MiniMax-M3.1-Flash
```

vLLM: reasoning not split from content when streaming (non-streaming works)

by mmssix - opened about 11 hours ago

Discussion

mmssix

about 11 hours ago

When serving MiniMax-M3.1-Flash with vLLM and --reasoning-parser minimax_m3, non-streaming responses split reasoning from content correctly, but streaming responses do not: the whole <mm:think>...</mm:think> block arrives in delta.content and delta.reasoning stays empty.

Environment

Image: vllm/vllm-openai:minimax-m31-flash (digest sha256:86f582dae25ea92802353727164ae4eb03b37b18baf9119b805ffaec2a7ef07c, vLLM 0.1.dev23341+g8744dd2fb)
Weights: this repo at main (54e4e69), chat_template.jinja unmodified
4x RTX PRO 6000 Blackwell (sm120), TP4

Relevant serve args:

--tensor-parallel-size 4 --block-size 128 --disable-hybrid-kv-cache-manager
--kv-cache-dtype fp8 --attention-config '{"indexer_kv_dtype":"fp8"}'
--reasoning-parser minimax_m3 --enable-auto-tool-choice --tool-call-parser minimax_m3

Repro

Same prompt both ways: {"messages":[{"role":"user","content":"What is 17*23? Answer briefly."}]}

Non-streaming (correct):

reasoning: "17*23 = 391."
content:   "391"

Streaming ("stream": true), concatenated deltas:

reasoning: ""
content:   "<mm:think>17*23 = 391.</mm:think>391"

The first content delta is literally "<mm:think>".

Reproduced directly against the vLLM server, with no proxy in between.

What we tried

Request	Streaming result
default (adaptive thinking)	not split, `<mm:think>...</mm:think>` in content
`"chat_template_kwargs": {"thinking_mode": "enabled"}`	not split
`thinking_mode: enabled` + `"reasoning_effort": "high"`	not split, and content begins with a stray `</mm:think>`

vllm/reasoning/minimax_m3_reasoning_parser.py has not changed upstream since 2026-08-31, so a newer nightly does not seem to address it.

Is this a known issue with the vLLM integration, or is there a serving flag or chat-template kwarg we are missing for streaming? Happy to test a fix.

Side note for other sm120 (RTX PRO 6000 / consumer Blackwell) users

The vLLM recipe targets SM100. On sm120:

--kv-cache-dtype nvfp4 and indexer_kv_dtype: nvfp4 are rejected (SM100/SM103/SM107 only), so we use fp8.
The DSpark draft needs an explicit attention_backend. Left on auto it picks FlashInfer, which has no 128-token page and fails with No common block size for 128. FLASH_ATTN cannot hold an fp8 KV cache on sm120, so we use "attention_backend":"FLASH_ATTN","kv_cache_dtype":"bfloat16" for the draft.

With that it loads and runs well: ~240 tok/s single-stream decode, DSpark mean acceptance length ~3.05 with 3 speculative tokens.

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