Rerank 3x faster than Cohere Rerank 4 Pro, 94% cheaper

SIE runs open-source reranking models behind one API. Send a question and the passages your search found, add a rule for what counts as relevant when you need one, and get the passages back best first.

Get started Run in your cloud

SIE Qwen3 Reranker 4B

Condition matched

Query

National Wildlife Refuge System; 2026-2027 Station-Specific Hunting and Sport Fishing Regulations

Relevance condition

Return the final rule the agency has adopted on this subject, not a proposed rule. A proposal that is still open for comment is not relevant.

1

Final rule1 Sept 2026

We, the U.S. Fish and Wildlife Service (FWS or Service), open or expand hunting opportunities on 111 field stations, including 107 units of the National Wildlife Refuges System…

0.995score
2

Proposal27 May 2026

0.946score

Cohere Rerank 4 Pro puts the wrong rule first

Query

National Bank Chartering

Relevance condition

Final rule on the requested subject

Cohere Rerank 4 Pro

Wrong source

Rank 1 · Final rule20 Dec 2016

Receiverships for Uninsured National Banks

The Office of the Comptroller of the Currency (OCC) is adopting a final rule addressing the conduct of receiverships for national banks that are not insured by the Federal Deposit Insurance Corporation (FDIC) (uninsured banks) and for which the FDIC would not be appointed…

Read official source ↗

arXiv:2605.05253 Study: retrieval over company documents with vector search alone answered 51% of questions correctly
EnterpriseRAG-Bench: A RAG Benchmark for Company Internal Knowledge

Yuhong Sun, Joachim Rahmfeld, Chris Weaver et al.
5 May 2026

3x faster than Cohere Rerank 4 Pro 94% cheaper, 1.6x faster than Voyage rerank‑3

Run this eval

Right document first

499 questions ↑

GPT-6 Luna (high reasoning)

GPT-6 Luna (high reasoning) 98.4%

Voyage rerank-3

Voyage rerank-3 92.6%

SIE Qwen3 Reranker 4B

SIE Qwen3 Reranker 4B 91.2%

Cohere Rerank 4 Pro

Cohere Rerank 4 Pro 84.2%

Claude Haiku 5.5 (thinking off)

Claude Haiku 5.5 (thinking off) 83.8%

Cohere Rerank 4 Fast

Cohere Rerank 4 Fast 71.5%

Time per query

median ↓

SIE Qwen3 Reranker 4B

55 ms SIE Qwen3 Reranker 4B

Voyage rerank-3

91 ms Voyage rerank-3

Cohere Rerank 4 Fast

105 ms Cohere Rerank 4 Fast

Cohere Rerank 4 Pro

179 ms Cohere Rerank 4 Pro

Claude Haiku 5.5 (thinking off)

Claude Haiku 5.5 (thinking off) 1.30 s

GPT-6 Luna (no reasoning)

GPT-6 Luna (no reasoning) 1.38 s

Cost per 100,000 queries

USD ↓

SIE Qwen3 Reranker 4B

$15 SIE Qwen3 Reranker 4B

Voyage rerank-3

$17 Voyage rerank-3

Claude Haiku 5.5 (thinking off)

$51 Claude Haiku 5.5 (thinking off)

GPT-6 Luna (high reasoning)

$86 GPT-6 Luna (high reasoning)

Cohere Rerank 4 Fast

Cohere Rerank 4 Fast $200

Cohere Rerank 4 Pro

Cohere Rerank 4 Pro $250

Tested: 499 Federal Register questions with 20 candidates each, under two relevance rules
Scored: A question is right when the requested document ranks first under both rules
Prices: Each model at its list price; SIE at its Serverless price
Timing: SIE timed on a GPU server with the client alongside; providers over their public APIs

Set your relevance condition in the request

View on GitHub

from sie_sdk import SIEClient
client = SIEClient(api_key="API keysk-sie-…")
query = "queryNational Wildlife Refuge System; 2026-2027 Station-Specific Hunting and Sport Fishing Regulations"
documents = [
    "document 1···ACTION: Final rule.···",
    "document 2···ACTION: Proposed rule.···",
]
items = [{"id": str(i), "text": d} for i, d in enumerate(documents)]
ranked = client.score("modelQwen/Qwen3-Reranker-4B", {"text": query}, items,
    instruction="instructionReturn the final rule the agency has adopted on this subject, not a proposed rule. A proposal that is still open for comment is not relevant.")
for r in ranked["scores"]:
    print(r["score"], documents[int(r["item_id"])])

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

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

const query = "National Wildlife Refuge System; 2026-2027 Station-Specific Hunting and Sport Fishing Regulations";
const documents = [
  "···ACTION: Final rule.···",
  "···ACTION: Proposed rule.···",
];
const items = documents.map((text, i) => ({ id: String(i), text }));
const ranked = await client.score('Qwen/Qwen3-Reranker-4B', { text: query }, items,
  { instruction: "Return the final rule the agency has adopted on this subject, not a proposed rule. A proposal that is still open for comment is not relevant." });
console.log(ranked.scores.map(({ itemId, score }) => ({ document: documents[Number(itemId)], score })));

curl https://api.superlinked.com/v1/score/Qwen%2FQwen3-Reranker-4B \
  -H "Authorization: Bearer sk-sie-…" \
  -H "Content-Type: application/json" \
  -d "{\"query\":{\"text\":\"National Wildlife Refuge System; 2026-2027 Station-Specific Hunting and Sport Fishing Regulations\"},\"items\":[{\"id\":\"0\",\"text\":\"···ACTION: Final rule.···\"},{\"id\":\"1\",\"text\":\"···ACTION: Proposed rule.···\"}],\"instruction\":\"Return the final rule the agency has adopted on this subject, not a proposed rule. A proposal that is still open for comment is not relevant.\"}"

Build the "Rerank" 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-Reranker-4B (SIE primitive: /score). Keep the model id configurable.

Task
- Input: a query string plus a list of candidate documents.
- Behaviour: reorder the candidates by true relevance to the query using the cross-encoder
- 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

RankSourceScore

Final rule

National Wildlife Refuge System; 2026–2027 Station-Specific Hunting and Sport Fishing Regulations

2026-17895 ↗

0.995

Proposal

National Wildlife Refuge System; 2026–2027 Station-Specific Hunting and Sport Fishing Regulations

2026-10465 ↗

0.946

Final rule ranked first

Rerank 100,000 queries for $15

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
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-default

Quickstart

Switch to SIE in 5 minutes

import cohere

client = cohere.ClientV2(api_key="YOUR_COHERE_API_KEY")
ranked = client.rerank(
    model="rerank-v4.0-pro", query=f"{condition}\n{query}",
    documents=documents,
)

import voyageai

client = voyageai.Client(api_key="YOUR_VOYAGE_API_KEY")
ranked = client.rerank(
    model="rerank-3", query=f"{condition}\n{query}",
    documents=documents,
)

SIE

from sie_sdk import SIEClient

client = SIEClient(api_key="YOUR_SIE_API_KEY")
ranked = client.score(
    "Qwen/Qwen3-Reranker-4B", {"text": query},
    [{"id": str(i), "text": doc} for i, doc in enumerate(documents)],
    instruction=condition,
)

from sie_sdk import SIEClient

client = SIEClient(api_key="YOUR_SIE_API_KEY")
ranked = client.score(
    "Qwen/Qwen3-Reranker-4B", {"text": query},
    [{"id": str(i), "text": doc} for i, doc in enumerate(documents)],
    instruction=condition,
)