Instructions to use JetBrains/Qwen3.6-27B-DFlash2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use JetBrains/Qwen3.6-27B-DFlash2 with Transformers:

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

pipe = pipeline("text-generation", model="JetBrains/Qwen3.6-27B-DFlash2")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("JetBrains/Qwen3.6-27B-DFlash2")
model = AutoModel.from_pretrained("JetBrains/Qwen3.6-27B-DFlash2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use JetBrains/Qwen3.6-27B-DFlash2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "JetBrains/Qwen3.6-27B-DFlash2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "JetBrains/Qwen3.6-27B-DFlash2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/JetBrains/Qwen3.6-27B-DFlash2

SGLang

How to use JetBrains/Qwen3.6-27B-DFlash2 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 "JetBrains/Qwen3.6-27B-DFlash2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "JetBrains/Qwen3.6-27B-DFlash2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "JetBrains/Qwen3.6-27B-DFlash2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "JetBrains/Qwen3.6-27B-DFlash2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use JetBrains/Qwen3.6-27B-DFlash2 with Docker Model Runner:
```
docker model run hf.co/JetBrains/Qwen3.6-27B-DFlash2
```

Qwen3.6-27B-DFlash2 / CHANGES.md

vvsotnikov

Add BF16 DFlash2 drafter for Qwen3.6-27B (thinking disabled)

a712ff7 verified 13 days ago

preview code

Raw History Blame Contribute Delete

1.27 kB

Changes for Qwen3.6-27B-DFlash2

First released as vvsotnikov/Qwen3.6-27B-nothink-DFlash2-v0.1; this repository publishes the same weights under the JetBrains organization. The weights are unchanged; only naming and documentation were updated.

Adapted the already-trained vvsotnikov/Qwen3.8-27B-test-DFlash2-v0.3 draft (weights SHA-256 fbee4cc4790035d0a4d4d7cece25f41a0f8d629b4c630d19be86a994ee90a477) for stock Qwen/Qwen3.6-27B with thinking disabled, using a fresh optimizer.
Released the BF16 inference export after 100 additional updates. model.safetensors contains the modified draft parameters; SHA-256: 52d968b9a9f7cc077580327074119a3815f74f48296c5f05b1aa1e0713f7b1f8.
Retained the DFlash2 architecture in config.json and added release-specific documentation. This distribution excludes optimizer state and training data.
Preserved the inherited Apache-2.0 license and attribution notices. The inherited NOTICE describes the prior blend adaptation; its additional modification section identifies this release's stock Qwen3.6 target.

The export was verified against its saved checkpoint. Serving performance and downstream quality have not been evaluated for this release.