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.
cohere/embed-5-proRuns on your balance at the prices above. New accounts start with free usage.
Settings
Cohere: Embed 5 Pro is a multimodal embedding model from Cohere, positioned as the higher-accuracy tier of the Embed 5 family. It is built for producing embeddings rather than generating text, and it sits in the embeddings category of models available through api14.
The model accepts text, images, and interleaved text and image content as input, and it returns embeddings as its output. It supports a context length of up to 128,000 tokens, with a maximum output of 115,200. Because it handles both modalities and combinations of them, the same model can embed a paragraph, a screenshot, or a document page where text and visuals appear together, which keeps retrieval pipelines from needing separate text and image embedding models.
The practical uses follow directly from that. Semantic search and retrieval-augmented generation over mixed corpora are the obvious fits, including knowledge bases where articles, diagrams, and scanned pages live side by side. Recommendation and similarity systems can use the embeddings to compare items across text and image representations. Clustering, deduplication, and classification over large document sets also work well, since the model can place text and images in a shared embedding space. Multimodal search, where a text query retrieves relevant images or a visual query retrieves relevant passages, is another natural application.
You call the model through api14's OpenAI-compatible API using the model id "cohere/embed-5-pro". Because the endpoint follows the OpenAI embeddings format, existing clients and SDKs that already point at an OpenAI-compatible base URL work without changes beyond the model id. The model does not list any supported request parameters, so requests rely on the standard embeddings inputs: the content to embed and the model id itself. There are no sampling or generation parameters to configure, since the model only produces embeddings.
If your task is text generation, chat, or any form of completion, this is not the right model, because its only output is embeddings. It is also worth checking whether a text-only embedding model is sufficient for your data; if your corpus contains no images and no interleaved content, the multimodal capability goes unused. For workloads that do involve images or mixed content, Embed 5 Pro is the higher-accuracy option within the Embed 5 family.
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api14.com/api/v1",
apiKey: process.env.API14_API_KEY,
});
const result = await client.embeddings.create({
"model": "cohere/embed-5-pro",
"input": [
"The quick brown fox",
"Lorem ipsum"
]
});
console.log(result.data.map((d) => d.embedding.length));
This community has no topics yet. Open the first one about Cohere: Embed 5 Pro.