Classify text 74x faster than GPT‑6 Luna, 85% cheaper

Send a message and the categories your app uses, such as support topics or next actions. SIE, an open-source inference engine, serves GLiNER2.5 Decide, an open classifier that scores every category in one call, so your code acts on the top one with no prompt to write and no reply to parse.

Get started Run in your cloud

SIE GLiNER2.5 Decide

Right category on top

Message

I've been charged more than once for the same transaction

Your categories Score

duplicate card payment 0.81
request a card payment refund 0.07
card payment fee 0.05
unrecognized card payment 0.02
reverted card payment 0.01
declined card payment 0.01
pending card payment 0.01
card not working 0.01

Median time per message 13.1 ms

A language model API makes your app wait 0.97 to 1.03 seconds for one category

Message

There's a recent charge on my card that I know I didn't make because I've never seen the name before. Can we investigate this?

GPT-6 Luna (no reasoning)

Over 100 ms

Category returned

unrecognized card payment

Median wait 0.97 s

100 ms to feel instant

Claude Haiku 5.5 (thinking off)

Over 100 ms

Category returned

unrecognized card payment

Median wait 1.03 s

100 ms to feel instant

Nielsen Norman Group Nielsen Norman Group: a reply feels instant only within 0.1 second
Response Times: The 3 Important Limits

“0.1 second is about the limit for having the user feel that the system is reacting instantaneously, meaning that no special feedback is necessary except to display the result.”
1 January 1993
Deloitte Deloitte: a mobile site 0.1 second faster lifted retail conversions by 8.4%
Milliseconds Make Millions

“Retail conversions increased by 8.4% and average order value increased by 9.2%.”
24 March 2020

74x faster than GPT‑6 Luna at 91% of its accuracy, 85% cheaper

Run this eval

Labelled correctly

400 messages ↑

GPT-6.1 Sol

GPT-6.1 Sol 97.3%

Claude Sonnet 5.5

Claude Sonnet 5.5 96.3%

GPT-6 Luna (high reasoning)

GPT-6 Luna (high reasoning) 95.8%

Claude Haiku 5.5 (thinking off)

Claude Haiku 5.5 (thinking off) 95.0%

GPT-6 Luna (no reasoning)

GPT-6 Luna (no reasoning) 93.8%

SIE GLiNER2.5 Decide

SIE GLiNER2.5 Decide 87.5%

Time per message

median ↓

SIE GLiNER2.5 Decide

13.1 ms SIE GLiNER2.5 Decide

GPT-6 Luna (no reasoning)

0.97 s GPT-6 Luna (no reasoning)

Claude Haiku 5.5 (thinking off)

1.03 s Claude Haiku 5.5 (thinking off)

Claude Sonnet 5.5

1.13 s Claude Sonnet 5.5

GPT-6 Luna (high reasoning)

1.35 s GPT-6 Luna (high reasoning)

GPT-6.1 Sol

1.48 s GPT-6.1 Sol

Cost per 1,000 messages

USD ↓

SIE GLiNER2.5 Decide

$0.0035 SIE GLiNER2.5 Decide

GPT-6 Luna (no reasoning)

$0.0239 GPT-6 Luna (no reasoning)

GPT-6 Luna (high reasoning)

$0.0372 GPT-6 Luna (high reasoning)

Claude Haiku 5.5 (thinking off)

$0.0486 Claude Haiku 5.5 (thinking off)

GPT-6.1 Sol

GPT-6.1 Sol $0.525

Claude Sonnet 5.5

Claude Sonnet 5.5 $1.04

Tested: 400 customer messages from public test sets, 8 categories each
Scored: A message is right when the model picks the dataset's category
Prices: OpenAI and Anthropic list prices; SIE at its Serverless price
Timing: SIE timed on a GPU server with the client alongside; providers over their public APIs

Send your categories with the message, get a score for each

from sie_sdk import SIEClient
client = SIEClient(api_key="sk-sie-…")
result = client.extract(
    "modelfastino/GLiNER2.5-Decide",
    {"text": "textI've been charged more than once for the same transaction"},
    labels=[
        "labelpending card payment",
        "labelreverted card payment",
        "labelrequest a card payment refund",
        "labelduplicate card payment",
        "labelcard payment fee",
        "labelunrecognized card payment",
        "labeldeclined card payment",
        "labelcard not working",
    ],
)
print(result["classifications"])

import { SIEClient } from '@superlinked/sie-sdk';

const client = new SIEClient(undefined, { apiKey: 'sk-sie-…' });

const result = await client.extract(
  'fastino/GLiNER2.5-Decide',
  { text: "I've been charged more than once for the same transaction" },
  { labels: ["pending card payment","reverted card payment","request a card payment refund","duplicate card payment","card payment fee","unrecognized card payment","declined card payment","card not working"] },
);
console.log(result.classifications);

curl https://api.superlinked.com/v1/extract/fastino%2FGLiNER2.5-Decide \
  -H "Authorization: Bearer sk-sie-…" \
  -H "Content-Type: application/json" \
  -d "{\"items\":[{\"text\":\"I've been charged more than once for the same transaction\"}],\"params\":{\"labels\":[\"pending card payment\",\"reverted card payment\",\"request a card payment refund\",\"duplicate card payment\",\"card payment fee\",\"unrecognized card payment\",\"declined card payment\",\"card not working\"]}}"

Build the "Classify text" 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: fastino/GLiNER2.5-Decide (SIE primitive: /extract). Keep the model id configurable.

Task
- Input: a block of text plus a set of candidate labels.
- Behaviour: return the best-matching label(s) with native scores
- 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.

Run

Output

Category Score

duplicate card payment 0.81
request a card payment refund 0.07
card payment fee 0.05
unrecognized card payment 0.02
reverted card payment 0.01
declined card payment 0.01
pending card payment 0.01
card not working 0.01

Run the classifier on Serverless or beside your app

Serverless API

We host the models; you call one API.

No idle GPUs, pay for what you use
Fits your stack: SDK, API, CLI, MCP
Self-host the same open-source engine
SOC 2 Type 2, US or EU data residency
Create an account to try Serverless

Get started

Self-hosted (Kubernetes)

Your data stays in your cloud, backed by our engineers.

Terraform modules for your cloud
Apache-2.0, same engine as Serverless
GPU workers scale to zero between jobs
Separate worker pools for each tenant

Agent prompt Deploy SIE to our AWS account with the superlinked/sie/aws Terraform module. Docs: superlinked.com/docs/deployment.

Deploy SIE to our Google Cloud project with the superlinked/sie/google Terraform module. Docs: superlinked.com/docs/deployment.

Deploy SIE to our Azure subscription with the superlinked/sie/azure Terraform module. Docs: superlinked.com/docs/deployment.

Deploy SIE to our Alibaba Cloud account with the superlinked/sie/alicloud Terraform module. Docs: superlinked.com/docs/deployment.

Talk to us Deploy guide

Run locally

The open-source server on your own machine.

NVIDIA GPU
One container, no cluster
Process your data fully offline
Same SDK and model IDs as Serverless

docker run --gpus all -p 8080:8080 ghcr.io/superlinked/sie-server:latest-cuda12-transformers5

Quickstart

Switch to SIE in 5 minutes

import json
from openai import OpenAI

client = OpenAI()
response = client.responses.create(
    model="gpt-6-luna",
    input=prompt,
    text={"format": schema},
)
reply = json.loads(response.output_text)
category = reply["category"]

from openai import OpenAI

client = OpenAI()
decision = client.decisions.create(
    model="gpt-6-luna",
    input=message,
    questions=[{
        "type": "choice",
        "instructions": "Which category fits this message?",
        "choices": [{"value": c} for c in categories],
    }],
)
category = decision.answers[0].choice

import json
import anthropic

client = anthropic.Anthropic()
reply = client.messages.create(
    model="claude-haiku-5-5",
    max_tokens=8192,
    messages=[{"role": "user", "content": prompt}],
    output_config={"format": schema},
)
category = json.loads(reply.content[0].text)["category"]

from typesafe_sdk import Choice, TypeSafeClient

client = TypeSafeClient(model="jev-latest")
response = client.system_one(
    state=message,
    questions={"category": Choice(
        instructions="Which category fits this message?",
        criteria=dict.fromkeys(categories),
    )},
)
category = response.choices["category"].choice

SIE

from sie_sdk import SIEClient

client = SIEClient(api_key="sk-sie-…")
result = client.extract(
    "fastino/GLiNER2.5-Decide",
    {"text": message},
    labels=categories,
)
top = result["classifications"][0]
category = top["label"]

from openai import OpenAI

client = OpenAI(
    api_key="sk-sie-…",
    base_url="https://api.superlinked.com/v1",
)
decision = client.decisions.create(
    model="fastino/GLiNER2.5-Decide",
    input=message,
    questions=[{
        "type": "choice",
        "instructions": "Which category fits this message?",
        "choices": [{"value": c} for c in categories],
    }],
)
category = decision.answers[0].choice

from sie_sdk import SIEClient

client = SIEClient(api_key="sk-sie-…")
result = client.extract(
    "fastino/GLiNER2.5-Decide",
    {"text": message},
    labels=categories,
)
top = result["classifications"][0]
category = top["label"]

from typesafe_sdk import Choice, TypeSafeClient

client = TypeSafeClient(
    api_key="sk-sie-…",
    base_url="https://api.superlinked.com",
    model="fastino/GLiNER2.5-Decide",
)
response = client.system_one(
    state=message,
    questions={"category": Choice(
        instructions="Which category fits this message?",
        criteria=dict.fromkeys(categories),
    )},
)
category = response.choices["category"].choice