Retrieval models

Pareta builds state-of-the-art rerankers, embeddings, and custom models for production AI systems — lightweight, blazing fast, and accurate where generalist models aren't. Two of them carry a retrieval stack: pareta-embed and pareta-rerank, each measured above its frontier baseline at a fraction of the price. Call them directly from any stack, or let Pareta use them for you.

pareta-embed

text embeddings · POST /v1/embeddings
109%of OpenAI text-embedding-3-large quality (nDCG@10)
32×cheaper than the same baseline
$0.004 / 1M input tokens

pareta-rerank

document reranking · POST /v1/rerank
101%of Gemini Flash reranking quality (nDCG@10)
31×cheaper than the same baseline
$0.025 / 1,000 documents

Use them directly

Standard REST endpoints with your Pareta API key — they work with any retrieval stack, no other Pareta usage required.

curl https://api.pareta.ai/v1/embeddings \
  -H "Authorization: Bearer $PARETA_API_KEY" \
  -d '{"input": ["governed metrics catalog", "..."]}'

curl https://api.pareta.ai/v1/rerank \
  -H "Authorization: Bearer $PARETA_API_KEY" \
  -d '{"query": "termination clause", "documents": ["...", "..."]}'

Or don't call them at all: requests to model:"auto" that need retrieval use the same models automatically.

Need dedicated capacity — or a model trained for your domain? Both models can be pinned to reserved GPUs for your workload — fixed latency, no cold starts, volume pricing — and we train custom rerankers and task models on your data. Write to .
Start free →
Quality measured on Pareta benchmark runs, July 2026, on the production serving path customers call. nDCG@10, normalized to the named baseline. Frontier cost = usage tokens × vendor list price; Pareta = actual metered billing.