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

Libraries

How to use nbeerbower/Wichtel-Qwen3.6-27B with Transformers:

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

pipe = pipeline("image-text-to-text", model="nbeerbower/Wichtel-Qwen3.6-27B")
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("nbeerbower/Wichtel-Qwen3.6-27B")
model = AutoModelForMultimodalLM.from_pretrained("nbeerbower/Wichtel-Qwen3.6-27B", 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 nbeerbower/Wichtel-Qwen3.6-27B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nbeerbower/Wichtel-Qwen3.6-27B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nbeerbower/Wichtel-Qwen3.6-27B",
		"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/nbeerbower/Wichtel-Qwen3.6-27B

SGLang

How to use nbeerbower/Wichtel-Qwen3.6-27B 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 "nbeerbower/Wichtel-Qwen3.6-27B" \
    --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": "nbeerbower/Wichtel-Qwen3.6-27B",
		"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 "nbeerbower/Wichtel-Qwen3.6-27B" \
        --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": "nbeerbower/Wichtel-Qwen3.6-27B",
		"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 nbeerbower/Wichtel-Qwen3.6-27B with Docker Model Runner:
```
docker model run hf.co/nbeerbower/Wichtel-Qwen3.6-27B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Wichtel-Qwen3.6-27B

Wichtel: the German house-imp — the little helper who quietly gets your jobs done, and knows which ones to hand to someone else.

A local operator model that scores 47/47 on a 47-case tool-use benchmark and writes Hemlock. Four adapters merged into one set of weights with no measurable interference between them.

Results

Tool use — 47/47, no misses

axis	cases	TIES	Vorarbeiter	Wichtel
delegate	10	0/10	7/10	10/10
no_tool (restraint)	10	9/10	9/10	10/10
trivial (one-line edits)	5	4/5	5/5	5/5
inspect	5	5/5	5/5	5/5
exec	4	4/4	4/4	4/4
git	4	4/4	4/4	4/4
memory	3	3/3	3/3	3/3
web	3	3/3	3/3	3/3
persona	3	3/3	3/3	3/3
total	47	35/47	41/47	47/47

47/47 valid tool calls, zero hallucinated tool names. Greedy decoding, so these are deterministic.

Hemlock — 67.6% on hembench

Every generated program is executed by the real interpreter and its stdout compared against expected output. Not string matching.

level		TIES	Schierling/TIES	Wichtel
L1	syntax	1/9	8/9	7/9
L2	stdlib	0/5	1/5	2/5
L3	algorithms	0/7	5/7	4/7
L4	systems	0/7	5/7	6/7
L5	translation	0/5	3/5	3/5
L6	debugging	3/5	4/5	4/5
overall	weighted	10.1%	65.0%	67.6%
	code that fails to run	30	8	7

Why this combination

No code agent understands Hemlock. Vorarbeiter's delegation policy hands engineering work to a coding agent, which is correct for every language that agent knows and wrong for this one — if the operator does not write Hemlock itself, nothing does. So this model has to do both: delegate ordinary engineering, and write Hemlock directly.

The interesting result is that the two capabilities did not fight. The risk was concrete: the Hemlock adapter is 5,562 rows of direct code generation with zero tool calls, exactly the data that could teach an operator to write code instead of delegating. Instead delegation went from 7/10 to 10/10.

An earlier attempt to get both from a single ORPO run failed in a specific, measurable way: folding 400 Hemlock preference pairs into the delegation data dropped restraint from 9/10 to 4/10, because all 400 rows call a tool and nothing in the training set teaches answering directly. Splitting the objectives — SFT for the language, preference training for the behaviour — fixed it and taught the language far better besides (400 preference pairs moved hembench approximately nothing; 5,562 supervised rows moved it 55 points).

Lineage

Qwen/Qwen3.6-27B
  └─ huihui-ai/Huihui-Qwen3.6-27B-abliterated
      └─ nbeerbower/Qwen3.6-27B-TIES                     5-way TIES merge
          └─ + Bubba-3ep            (GreatFirewall ORPO — states facts it otherwise sanitises)
              └─ + delegation LoRA  (hands engineering work to a code agent)  = Vorarbeiter
                  └─ + Schierling   (Hemlock SFT, 5562 rows)                  = Hemlock-Qwen3.6-27B
                      └─ + egirl LoRA (delegation + Hemlock tool use, ORPO)   ← Wichtel

Four adapters deep, all four capabilities intact.

MTP head

peft merge_and_unload() silently drops the 15 mtp.* tensors and current llama.cpp then refuses to load the model. They are grafted back from the pre-merge base and verified bit-identical. 1,199 tensors total (850 language, 333 vision, 15 MTP, 1 lm_head).

llama-server -m Wichtel-Q8_0.gguf -ngl 99 -c 8192 --jinja -fa on -np 1 \
    --spec-type draft-mtp --spec-draft-n-max 4

MTP speculation is not bit-identical in practice — on this bench one decision flipped between speculative and non-speculative decoding at temperature 0, apparently from numerical differences between batched verification and sequential decoding flipping a near-tie.

Limitations

47/47 is a saturated benchmark. It cannot distinguish this model from a better one, and 47 cases is a small sample — treat it as "no remaining failures we can see", not as a ceiling. The bench was widened from 10 cases to 47 during this work precisely because the 10-case version could not resolve the differences being claimed with it.
stdlib is the weak Hemlock axis (2/5). Standard-library work needs exact function names and signatures — memorised API knowledge rather than transferable syntax. Check calls against docs.
Single-turn tool use. It learns the first action, not what to do with results or how to recover from a failed call.
Delegation is bound to egirl's tool schema (code_agent, glob_files, git_status…). A different tool set needs different training data.
The Chinese-language censorship gap persists from the Bubba layer: asking 六四事件是什么? in Chinese remains the weakest item.
Prose is untested here. A sibling experiment found that a prose adapter trained on this family's own outputs degraded the model it was trained against — see Obscura.

nbeerbower/Hemlock-Qwen3.6-27B — this model without the egirl adapter
nbeerbower/Schierling-Qwen3.6-27B-LoRA — the Hemlock adapter
nbeerbower/Qwen3.6-27B-egirl-LoRA — the tool-use adapter
nbeerbower/Vorarbeiter-Qwen3.6-27B — the operator base
egirl — the agent this serves
hemlock — the language and its benchmark

Downloads last month: 53

Safetensors

Model size

27B params

Tensor type

BF16

Model tree for nbeerbower/Wichtel-Qwen3.6-27B

hemlang/Hemlock-Qwen3.6-27B

nbeerbower/Qwen3.6-27B-egirl-LoRA

Merge model

this model

Quantizations

2 models

nbeerbower
/

Wichtel-Qwen3.6-27B