Instructions to use GestaltLabs/Ornstein3.8-27B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use GestaltLabs/Ornstein3.8-27B with Transformers:

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

pipe = pipeline("image-text-to-text", model="GestaltLabs/Ornstein3.8-27B")
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("GestaltLabs/Ornstein3.8-27B")
model = AutoModelForMultimodalLM.from_pretrained("GestaltLabs/Ornstein3.8-27B", 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 GestaltLabs/Ornstein3.8-27B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "GestaltLabs/Ornstein3.8-27B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "GestaltLabs/Ornstein3.8-27B",
		"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/GestaltLabs/Ornstein3.8-27B

SGLang

How to use GestaltLabs/Ornstein3.8-27B 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 "GestaltLabs/Ornstein3.8-27B" \
    --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": "GestaltLabs/Ornstein3.8-27B",
		"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 "GestaltLabs/Ornstein3.8-27B" \
        --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": "GestaltLabs/Ornstein3.8-27B",
		"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 GestaltLabs/Ornstein3.8-27B with Docker Model Runner:
```
docker model run hf.co/GestaltLabs/Ornstein3.8-27B
```

DJLougen commited on 5 days ago

Commit

cfd3173

verified ·

1 Parent(s): e443c86

Add Qwen vs Ornstein GSM8K/GPQA comparison table

Browse files

Files changed (1) hide show

README.md +17 -8

README.md CHANGED Viewed

@@ -23,6 +23,17 @@ model-index:
     - name: accuracy
       type: accuracy
       value: 96.51
 ---
 ![Ornstein3.8-27B](ornstein3.8-27b.jpg)
@@ -41,15 +52,13 @@ This checkpoint injects **Ornstein thinking** into Qwen3.8-27B. It is an early m
 Qwen3.8-27B achieves an estimated **97.0%** accuracy on the full GSM8K benchmark when running in standard unquantized precision (BF16/FP8).
-| | |
-|---|---|
-| Score | **1273 / 1319 (96.51%)** |
-| Dataset | official GSM8K test (`test.jsonl`) |
-| Precision | BF16, Fireworks dedicated H100 |
-| Decoding | greedy (`temperature=0`, `top_k=40`, `max_tokens=4000`) |
-| Extraction | `message.content` first (`####` / `\\boxed{}` / last number) |
-This is a single-run estimate on this checkpoint, not a harness-averaged leaderboard number. It does not apply to GGUF quants.
 ## Support this work

     - name: accuracy
       type: accuracy
       value: 96.51
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: GPQA Diamond
+      type: gpqa
+      split: diamond
+    metrics:
+    - name: accuracy
+      type: accuracy
+      value: 64.65
 ---
 ![Ornstein3.8-27B](ornstein3.8-27b.jpg)
 Qwen3.8-27B achieves an estimated **97.0%** accuracy on the full GSM8K benchmark when running in standard unquantized precision (BF16/FP8).
+| Benchmark | Qwen3.8-27B (reported) | Ornstein3.8-27B (this run) |
+|---|---|---|
+| GSM8K | — | **96.51** (1273/1319; est. 97.0) |
+| GPQA Diamond | **89.2** | 64.65 (128/198) |
+Qwen figures are from the [Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B) card (vendor-reported; GSM8K is not published there). Ornstein figures are a single greedy BF16 run on a Fireworks dedicated H100 (`temperature=0`, `top_k=40`, `max_tokens=4000`), answers taken from `message.content` first. This is not a harness-matched reproduction of Qwen's GPQA setup, and it does not apply to GGUF quants.
 ## Support this work