Glance describes an image in plain English, from a one-line alt text to a detailed paragraph. Send a question instead and it answers about the image: what's the total on this receipt, what color is the car, is anyone wearing a helmet.
It's a small vision-language model, so treat answers as a first pass: it reads text and names things well, and it's weaker at counting and fine detail.
Use it
curl https://labs.greatwork.company/v1/run \
-H "Authorization: Bearer $GWL_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "glance-1",
"input": "https://labs.greatwork.company/examples/product.jpg"
}'const res = await fetch("https://labs.greatwork.company/v1/run", {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.GWL_API_KEY}`,
"Content-Type": "application/json",
},
body: JSON.stringify({
"model": "glance-1",
"input": "https://labs.greatwork.company/examples/product.jpg"
}),
});
const { output, usage } = await res.json();import os, requests
res = requests.post(
"https://labs.greatwork.company/v1/run",
headers={"Authorization": f"Bearer {os.environ['GWL_API_KEY']}"},
json={
"model": "glance-1",
"input": "https://labs.greatwork.company/examples/product.jpg"
},
)
output = res.json()["output"]# Send a local file instead of a URL
curl https://labs.greatwork.company/v1/run \
-H "Authorization: Bearer $GWL_API_KEY" \
-F model=glance-1 \
-F input=@photo.jpgSet GWL_API_KEY to a key from your dashboard. The bare name glance works too and follows new versions.
Options
input: An image: an https URL, a data URL (`data:image/png;base64,...`), or a file in a multipart upload. JPEG, PNG, WebP, GIF, AVIF or TIFF, up to 10 MB.
| Name | Type | Default | What it does |
|---|---|---|---|
question | string | Ask something about the image. Without it, Glance writes a caption. | |
detail | string short | medium | long | "short" | Caption length: short for alt text, long for a full description. Ignored when question is set. |
Output
| Field | Type | Meaning |
|---|---|---|
caption | string | The description (when no question is asked). |
answer | string | The answer (when question is set). |
question | string | Your question, echoed. |
Every response wraps this in {"id", "model", "output", "usage": {"units", "unit", "cost_usd"}, "latency_ms"}.
Examples
Alt text
{
"model": "glance-1",
"input": "/examples/product.jpg"
}{
"caption": "A turquoise colored water bottle is on a light brown wooden table.",
"width": 1024,
"height": 1024
}Captured on our production servers: 4210 ms, 1 image, $0.0015.
Ask a receipt
{
"model": "glance-1",
"input": "/examples/receipt.png",
"question": "What is the total on this receipt?"
}{
"question": "What is the total on this receipt?",
"answer": "The total on this receipt is $24.65.",
"width": 560,
"height": 760
}Captured on our production servers: 3525 ms, 1 image, $0.0015.
Describe a scene
{
"model": "glance-1",
"input": "/examples/street.jpg",
"detail": "medium"
}{
"caption": "In this image I can see a bicycle parked on the road. I can see a dog and a car. I can see buildings and the sky.",
"width": 1024,
"height": 1024
}Captured on our production servers: 4050 ms, 1 image, $0.0015.
Good for
- Alt text for CMS and social uploads
- Answer questions about receipts, labels and screenshots
- Describe product photos for search
- Accessibility for user-generated images
Limits and caveats
- Counting is unreliable (it may say two dogs when there's one).
- Takes 3 to 6 seconds per image.
- English only.
Pricing
$1.50 per 1,000 images. One unit per image. Failed calls aren't charged, and a call never costs more than its estimate (POST /v1/estimate tells you before you run it).
Built on
HuggingFaceTB/SmolVLM-256M-Instruct, 256M parameters. License: Apache-2.0. We run it on our servers.