Instructions to use FINAL-Bench/Darwin-35B-A3B-Opus with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use FINAL-Bench/Darwin-35B-A3B-Opus with Transformers:

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

pipe = pipeline("image-text-to-text", model="FINAL-Bench/Darwin-35B-A3B-Opus")
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 AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("FINAL-Bench/Darwin-35B-A3B-Opus")
model = AutoModelForMultimodalLM.from_pretrained("FINAL-Bench/Darwin-35B-A3B-Opus", 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=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use FINAL-Bench/Darwin-35B-A3B-Opus with vLLM:

Install from pip and serve model

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

SGLang

How to use FINAL-Bench/Darwin-35B-A3B-Opus 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 "FINAL-Bench/Darwin-35B-A3B-Opus" \
    --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": "FINAL-Bench/Darwin-35B-A3B-Opus",
		"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 "FINAL-Bench/Darwin-35B-A3B-Opus" \
        --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": "FINAL-Bench/Darwin-35B-A3B-Opus",
		"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 FINAL-Bench/Darwin-35B-A3B-Opus with Docker Model Runner:
```
docker model run hf.co/FINAL-Bench/Darwin-35B-A3B-Opus
```

Please release the Darwin model based on Qwen 3.5 with the updated Qwen 3.6 architecture. We're eagerly awaiting this version. Thank you!

by Mithnick - opened Apr 16

Discussion

Mithnick

Apr 16

Please release the Darwin model based on Qwen 3.5 with the updated Qwen 3.6 architecture. We're eagerly awaiting this version. Thank you!

SeaWolf-AI

FINAL_Bench org Apr 17

Please release the Darwin model based on Qwen 3.5 with the updated Qwen 3.6 architecture. We're eagerly awaiting this version. Thank you!

Thanks for the interest. We’ve already started rebuilding it with the 3.6 architecture.

Zouna

Apr 20

Have you taken a look at the APEX quantization option? I wonder if you could provide the darwin-qwen3.6 with an APEX option, as the model size is reduced significantly, but it retains most of its capabilities.

SeaWolf-AI

FINAL_Bench org Apr 21

Have you taken a look at the APEX quantization option? I wonder if you could provide the darwin-qwen3.6 with an APEX option, as the model size is reduced significantly, but it retains most of its capabilities.

Thanks for the suggestion — great point!

Yes, we've been looking at APEX quantization for MoE
models, and since Darwin-35B-A3B-Opus is built on the
Qwen3.5 MoE architecture that APEX has been benchmarked
on (showing remarkable results — even beating F16
perplexity in some cases), it's a natural fit.

We're planning to release APEX variants in the following
tiers over the next 1-2 weeks:
• APEX Quality (21 GB) — maximum accuracy
• APEX Balanced (22 GB) — Q8_0 quality, smaller
• APEX I-Quality — diverse imatrix calibration
• APEX Mini (~12 GB) — consumer 16GB VRAM friendly

Will announce once uploaded. Thanks for bringing this
up — it's exactly the kind of feedback that helps us
prioritize community needs.

(Note: you mentioned "darwin-qwen3.6" — our current
Darwin MoE series is built on Qwen3.5-A3B; we'll note
this naming convention clearly in the APEX release.)

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