relia-1

Relevance / Rerank Model wallet: … requests left

Score how relevant each article is to an entity. 1 request = 1 entity + up to 100 articles → billed 1 credit. API docs →

Know which news is actually relevant

relia-1 is a purpose-built relevance model for financial news. Give it an entity and a batch of articles, and it scores how relevant each one is to that entity — so you can filter the noise before it ever reaches your pipeline, dashboard, or LLM.

86.5%

Exact-bin agreement with a 3-LLM consensus

$0.01

To score 1,000 articles — up to 500× cheaper than LLM APIs

60%

Of the noise cut before it ever reaches your pipeline

The problem

News feeds are flooded with articles that only mention a ticker in passing. Deciding what's actually material is slow by hand and expensive with LLMs.

What it returns

Per article: a 0–1 relevance score, an argmax level (0 irrelevant → 3 highly relevant), and the full probability distribution — ready to threshold or rank.

Why Relia

Purpose-trained for finance, deterministic, no prompt engineering. One flat price per request of up to 100 articles — long articles are never re-chunked.

Try it live

Input

1 Which company to score against?

relia scores how relevant each article below is to this company.

2 Articles to score for relevance to TSLA

Pre-filled with example articles — edit them or add your own.

Output