We use cookies. Cookies and browser storage keep you signed in and remember your preferences. With your permission we also measure how the site is used. You can any time. See our cookie policy.
mongodb/rerank-2.5-liteRuns on your balance at the prices above. New accounts start with free usage.
Settings
VoyageAI by MongoDB: rerank-2.5-lite is a reranking model from VoyageAI by MongoDB. It sits in the rerank category, which means its job is to take a set of candidate documents and reorder them by relevance to a query rather than to generate new text. It was released by the provider on 2026-07-27.
The model accepts text as input and produces a rerank output, so you send it a query together with the documents you want scored and it returns them in a more useful order. It has a context length of 32000 tokens and a maximum output of 28800. Because it works on text only, any candidate passages you want ranked need to be supplied as text. The provider describes it as a reranker optimized for both latency and quality, and reports a 7.16% improvement in retrieval accuracy over Cohere Rerank v3.5 across 93 datasets, along with outperforming Cohere Rerank v3.5.
The practical fit follows from that. If you already run a retrieval step, whether vector search, keyword search or a hybrid of the two, this model can sit after it and reorder the top candidates before they reach your application or your language model. That helps when the first stage returns plausible but imperfectly ordered results, which is common in search over documentation, support tickets, product catalogues or internal knowledge bases. It is also useful in retrieval augmented generation pipelines, where the quality of the passages placed near the top of the prompt has a direct effect on the answer, and in evaluation work where you want a consistent relevance ordering across a fixed set of candidates. The emphasis on latency makes it a reasonable choice when reranking happens on the request path rather than in a batch job.
You call it through api14's OpenAI-compatible API using the model id "mongodb/rerank-2.5-lite". Because the endpoint follows the OpenAI format, you can point an existing OpenAI client at api14 by changing the base URL and supplying your api14 key, then pass the model id in your request. No additional request parameters are supported for this model, so keep your calls to the standard fields the endpoint accepts rather than relying on provider specific options such as custom truncation or instruction settings.
Keep the context length in mind when assembling inputs: the query and the documents you submit share the 32000 token budget, so very large candidate sets need to be trimmed or split before you send them. If your task is generating text rather than ordering it, a chat or completion model is the right tool instead, and if you need to rank images or other non-text inputs, this model does not cover that.
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api14.com/api/v1",
apiKey: process.env.API14_API_KEY,
});
const result = await client.post("/rerank", { body: {
"model": "mongodb/rerank-2.5-lite",
"query": "What is the capital of France?",
"documents": [
"Paris is the capital of France.",
"Berlin is in Germany."
],
"top_n": 1
} });
console.log(result.results);
This community has no topics yet. Open the first one about VoyageAI by MongoDB: rerank-2.5-lite.