Why did we open-source our inference engine? Read the post

Let your agent read the photo before it acts

Superlinked gives your agent one API to ask an equipment photo or a wiring diagram a specific question, such as whether a breaker is locked out.

Get started
Siemens main switch in a panel door, secured with a padlock and a warning tag
Question

Can a technician switch this main breaker on right now? Answer yes or no, then say what in the photo decides it.

Qwen3.8-27B-FP8

No, the technician cannot switch this main breaker on right now because it is secured with a padlock and has a red tag attached, indicating it is intentionally locked out for safety or maintenance purposes.

Caption: locked and tagged NE VKLAPLJAJ, do not switch on

Placards and wiring diagrams come back exactly right

View evidence
Piping and instrumentation diagram showing a reactor vessel, a drain tank and a relief setting
Reactor experiment piping and instrumentation diagram

What is the drain tank's capacity, and what pressure is its relief device set at?

The drain tank has a capacity of 6000 gallons, and its relief device is set at 50 psig.

Matches Drawing: 6000 GAL, KD #2 SET AT 50 PSIG

U.S. Atomic Energy Commission, OSTI 4307822 · public domain
Wiring diagram of an apartment building distribution board with breaker ratings and cable types
Apartment building distribution board diagram

Which circuits in this wiring diagram are protected by B6 breakers, and what load does each one carry?

The circuits protected by B6 breakers are the "Stairs I lighting" and "Stairs II lighting" circuits, each carrying a load of P=120W.

Matches Drawing: Stairs I and Stairs II lighting, B6, 120 W each

Dmitry G via Wikimedia Commons · public domain
Rear of a fuel service truck carrying a diamond hazardous materials placard
Fuel truck hazmat placard

What UN number is on this truck's hazmat placard, and what hazard class does the placard show?

The UN number on the truck's hazmat placard is 1993, and it indicates a Class 3 flammable liquid hazard.

Matches Caption: placard 1993, diesel fuel

Ben Ostrowsky via Wikimedia Commons · CC BY 2.0
Absolute pressure gauge from the Mir space station with overlapping millimetre scales
Mir station absolute pressure gauge

What pressure does this gauge read, in millimetres of mercury?

Based on the position of the needle, this gauge reads approximately **130 mm Hg**.

Miss Page author's reading: about 761 mm Hg

Steve Jurvetson via Wikimedia Commons · CC BY 2.0

These 5 come from 12 recorded questions, and 10 answers matched the published caption, the drawing or the display. Both misses are here, and both read a needle against an analog dial.

Ask your own question about a photo and get a one-sentence answer

View on GitHub
from pathlib import Path
from sie_sdk import SIEClient
client = SIEClient(
api_key="API keysk-sie-…",
base_url="https://api.superlinked.com",
)
image = {"data": Path(imagetire-pressure-display.jpgbrowse).read_bytes(), "format": "jpeg"}
result = client.generate(
"modelQwen/Qwen3.8-27B-FP8",
"queryWhich tire has the lowest pressure on this display, and what is it? Start your reply with the answer in one sentence.",
max_new_tokens=256,
images=[image],
)
print(result["text"])
import { readFile } from 'node:fs/promises';
import { SIEClient } from '@superlinked/sie-sdk';

const client = new SIEClient('https://api.superlinked.com', {
  apiKey: 'sk-sie-…',
});

const image = await readFile("tire-pressure-display.jpg");
const result = await client.generate(
  'Qwen/Qwen3.8-27B-FP8',
  "Which tire has the lowest pressure on this display, and what is it? Start your reply with the answer in one sentence.",
  { maxNewTokens: 256, images: [image] },
);
console.log(result.text);
image_bytes=$(base64 < 'tire-pressure-display.jpg' | tr -d '\n')
curl https://api.superlinked.com/v1/generate/Qwen__Qwen3.8-27B-FP8 \
  -H "Authorization: Bearer sk-sie-…" \
  -H "Content-Type: application/json" \
  -d "{\"prompt\":\"Which tire has the lowest pressure on this display, and what is it? Start your reply with the answer in one sentence.\",\"max_new_tokens\":256,\"images\":[{\"data\":\"$image_bytes\",\"format\":\"jpeg\"}]}"
Build the "Caption / VQA" capability into my app using the Superlinked Inference Engine (SIE).

Context
- SIE is an OpenAI-style inference API. Python SDK: `from sie_sdk import SIEClient`; TypeScript: `@superlinked/sie-sdk`.
- Base URL: https://api.superlinked.com (or my regional endpoint). Auth: Bearer key from env `SIE_API_KEY` (never hard-code it).
- Model: Qwen/Qwen3.8-27B-FP8 (SIE primitive: /generate). Keep the model id configurable.

Task
- Input: an uploaded image plus a question or caption instruction.
- Behaviour: return a detailed caption or answer the question about the image
- Call the selected SIE primitive once per request and map the response into your domain type.

Deliverables
- A typed client wrapper, an application-level function for this task, error handling for timeouts/empty input, and unit tests with a stubbed client.
- Wire it into my existing stack (ask me which framework if unclear) and add a short usage example.
Output
Source tire-pressure-display.jpg
source · tire-pressure-display.jpg
The front right tire has the lowest pressure at 30 PSI.
Qwen3.8-27B-FP8

Image question answering quality and latency

PRICE
$ / 1M input tokens $ / 1M output tokens
$0.2 OpenAI GPT-5.4 nano
$0.3 Google Gemini 3.5 Flash-Lite
SIE Qwen3.5 4B $0.72
SIE Qwen3.6 27B $0.72
OpenAI GPT-5.4 mini $0.75
Anthropic Claude Haiku 4.5 $1
$0.72 SIE Qwen3.5 4B
$0.72 SIE Qwen3.6 27B
$1.25 OpenAI GPT-5.4 nano
$2.5 Google Gemini 3.5 Flash-Lite
OpenAI GPT-5.4 mini $4.5
Anthropic Claude Haiku 4.5 $5
QUALITY
VQAv2
SIE Qwen3.6 27B 0.71
0.69
0.68
SIE Qwen3.5 4B 0.62
0.60
0.59
LATENCY
p50 s
0.8s SIE Qwen3.5 4B
0.9s
1.0s
SIE Qwen3.6 27B 1.6s
1.8s
1.9s

Deploy your way

Self-host with K8s

Easy & scalable deployment in your own cloud.

  • Terraform to your cloud in minutes
  • Apache-2.0, same engine as Cloud
  • Scales to zero, no bill between jobs
  • Per-tenant pools, no noisy neighbors
Agent prompt
Deploy SIE to our AWS account with the superlinked/sie/aws Terraform module. Docs: superlinked.com/docs/deploymentDeploy SIE to our GCP project with the superlinked/sie/google Terraform module. Docs: superlinked.com/docs/deploymentDeploy SIE to our Azure AKS cluster via helm install. Requirements: superlinked.com/docs/deployment
Deploy guide

Run locally

Run the same models on your own machine.

  • Runs on NVIDIA GPU or Apple Silicon
  • One command, no Docker or cluster
  • All 100+ Cloud models, fully offline
  • Same SDK and IDs, no code changes
pip install "sie-server[local]" && sie-server servepip install "sie-server[local]" && sie-server serve --device cuda
Quickstart

Contact us

Tell us about your use case and we'll get back to you shortly.

Apply for an inference grant

Free capacity on our hosted cluster for selected projects. Tell us what you run and we reply by email.