Instructions to use omicverse/OmicAI-27B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use omicverse/OmicAI-27B with Transformers:

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

pipe = pipeline("text-generation", model="omicverse/OmicAI-27B")
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("omicverse/OmicAI-27B")
model = AutoModelForMultimodalLM.from_pretrained("omicverse/OmicAI-27B", 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 omicverse/OmicAI-27B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "omicverse/OmicAI-27B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "omicverse/OmicAI-27B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/omicverse/OmicAI-27B

SGLang

How to use omicverse/OmicAI-27B 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 "omicverse/OmicAI-27B" \
    --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": "omicverse/OmicAI-27B",
		"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 "omicverse/OmicAI-27B" \
        --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": "omicverse/OmicAI-27B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use omicverse/OmicAI-27B with Docker Model Runner:
```
docker model run hf.co/omicverse/OmicAI-27B
```

Request access to OmicAI-27B

Access to OmicAI-27B is reviewed manually and is free for academic and non-profit research — please apply using your institutional email address. Commercial use requires a separate license; email contact@omicos.cn with the subject line "OmicAI-27B commercial license". Please fill in the form as completely as you can, it is what we use to review the request.

OmicAI-27B

OmicAI is a family of language models optimized for omics data analysis and tool interaction over the omicverse toolkit. This is the 27B member — the largest released variant.

OmicAI-27B is Qwen3.8-27B fine-tuned with ms-swift / Together using LoRA (all-linear, rank 16, α 32) merged into the final weights.

The OmicAI family

Model	Base	Access
OmicAI-0.8B	Qwen3.5-0.8B	Open — Apache-2.0
OmicAI-4B	Qwen3.5-4B	Open — Apache-2.0
OmicAI-9B	MiMo-V2.6-Distill-Qwen-9B	Open — Apache-2.0
OmicAI-27B (this model)	Qwen3.8-27B	By application

Training Data

The training trajectories are produced by OmicOS and OmicVerse through a paper-reproduction process over the last three years of publications: the systems retrieve recent omics papers, then attempt to reproduce their analyses end-to-end from the raw data, yielding multi-turn agent trajectories (retrieval → code → omicverse tool calls → recovery from execution feedback). A trajectory is kept only when an automatic judge confirms the deliverable is legitimate and a scoreboard confirms the reproduced result matches or surpasses the paper's reported (SOTA) numbers.

Key Details


Base model	Qwen/Qwen3.8-27B
Method	LoRA (`all-linear`, rank 16, α 32, dropout 0.05) → merged; 1 epoch
Precision	bfloat16 (merged weights, 18 shards)
Chat template	`qwen3_5` — inference template must match training

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("omicverse/OmicAI-27B")
model = AutoModelForCausalLM.from_pretrained(
    "omicverse/OmicAI-27B", torch_dtype="bfloat16", device_map="auto")

Also loadable with vLLM (use the qwen3_5 chat template).

Intended Use & Limitations

Intended: research assistant for single-cell / bulk / spatial omics analysis and omicverse pipeline generation.
Limitations: a research model — generated analysis code should be executed and verified before use; may hallucinate APIs outside the omicverse/scanpy surface it was trained on.

Ecosystem

Part of the omicOS / omicAI effort and the omicverse ecosystem. Smaller members: OmicAI-4B · OmicAI-9B · OmicAI-0.8B.

Downloads last month: -

Safetensors

Model size

28B params

Tensor type

BF16

Model tree for omicverse/OmicAI-27B

Base model

Qwen/Qwen3.8-27B

Finetuned

(395)

this model