Instructions to use MIRA-MMDeepResearch/MIRA-SFT-35B-A3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MIRA-MMDeepResearch/MIRA-SFT-35B-A3B with Transformers:

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

pipe = pipeline("image-text-to-text", model="MIRA-MMDeepResearch/MIRA-SFT-35B-A3B")
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("MIRA-MMDeepResearch/MIRA-SFT-35B-A3B")
model = AutoModelForMultimodalLM.from_pretrained("MIRA-MMDeepResearch/MIRA-SFT-35B-A3B", 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 MIRA-MMDeepResearch/MIRA-SFT-35B-A3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MIRA-MMDeepResearch/MIRA-SFT-35B-A3B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MIRA-MMDeepResearch/MIRA-SFT-35B-A3B",
		"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/MIRA-MMDeepResearch/MIRA-SFT-35B-A3B

SGLang

How to use MIRA-MMDeepResearch/MIRA-SFT-35B-A3B 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 "MIRA-MMDeepResearch/MIRA-SFT-35B-A3B" \
    --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": "MIRA-MMDeepResearch/MIRA-SFT-35B-A3B",
		"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 "MIRA-MMDeepResearch/MIRA-SFT-35B-A3B" \
        --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": "MIRA-MMDeepResearch/MIRA-SFT-35B-A3B",
		"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 MIRA-MMDeepResearch/MIRA-SFT-35B-A3B with Docker Model Runner:
```
docker model run hf.co/MIRA-MMDeepResearch/MIRA-SFT-35B-A3B
```

MIRA-SFT-35B-A3B (MIRA-SFT)

MIRA is a multimodal deep research agent from the paper MIRA: A Multimodal Illustrated Deep Research Agent for Controllable Reports. This checkpoint is the agent model: it searches the web, opens pages and images, runs Python, and writes either a short answer or an illustrated report whose length, image count and source count follow the request. It is meant to run inside the MIRA agent runtime and harness (https://github.com/MIRA-MMDeepResearch/MIRA), which provide the tools, the report contract and the submission gate.


Base model	Qwen/Qwen3.6-35B-A3B
Architecture	`Qwen3_5MoeForConditionalGeneration`: 35B total / 3B active parameters, 256 experts (8 active), 40 layers, vision encoder
Context	262,144 tokens
Training	Full-parameter SFT in three stages on MIRA-Data (harness reports, multimodal search trajectories, failure-case repair), learning rate 5e-6, global batch 64
Precision	bfloat16, 70.2 GB in 2 safetensors shards

Use

Requires transformers>=5.2. Serve with vLLM and call it through the MIRA runtime:

vllm serve MIRA-MMDeepResearch/MIRA-SFT-35B-A3B --served-model-name MIRA \
  --max-model-len 262144 --enable-auto-tool-choice --tool-call-parser qwen3_coder \
  --allowed-local-media-path / --port 8000

python -m mira.harness.run --endpoint http://localhost:8000 --prompt-profile student \
  --mode detail --question "..." --output runs/example

Sampling. temperature 1.0, top-p 0.95, top-k 20, presence penalty 1.5 (the values in generation_config.json, except presence penalty, which the runtime sets).

Chat template. The bundled chat_template.jinja is the base model's template. Tools are described in the system turn and called with the XML format <tool_call><function=NAME><parameter=KEY>VALUE</parameter></function></tool_call>. Reasoning is on: keep enable_thinking true and pass preserve_thinking: true through chat_template_kwargs so earlier reasoning stays in context across tool calls. The system prompt the model was trained with is SYS_QWEN36_SFT_S2_MM in mira/agent/prompts/prompt.py; other system prompts reduce accuracy.

Evaluation

Accuracy (%) under the MIRA harness with the same judge for all agents. MMDeepResearch is the official report score.

Model	Params	MMSearch	MMSearch+	BrowseComp-VL	SimpleVQA	LiveVQA	FVQA	VDR-Bench	MMDeepResearch
OpenSearch-VL	30B (3B)	68.70	28.94	41.10	74.90	67.40	73.20	33.50	23.21
Vision-DeepResearch	30B (3B)	69.60	28.50	53.70	66.33	77.60	74.20	37.80	12.78
WeAgent-MMSearch	30B (3B)	77.78	37.82	–	77.73	81.05	79.33	35.68	–
Qwen3.6-35B-A3B + MIRA-Harness	35B (3B)	73.10	33.44	56.64	74.33	78.33	70.00	19.00	41.90
MIRA-SFT-35B-A3B	35B (3B)	80.70	38.91	58.40	79.00	82.00	79.33	29.80	42.61

"–" means not reported. Scores for other agents are from the MIRA paper.

Limitations

The report gate checks structure (length, images, sources, citations that resolve), not whether each claim is true. Web content changes over time, so scores drift. The model was distilled from teacher trajectories and inherits their errors.

Citation

@article{mira2027,
  title   = {MIRA: A Multimodal Illustrated Deep Research Agent for Controllable Reports},
  author  = {Wu, Hongrui and Tuo, Kaiwen and Wang, Congcong and Li, Gen and Dong, Shuai and
             Yu, Xinlei and Hou, Haowen and Peng, Zelin and Wang, Jiaqi},
  year    = {2027}
}