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-fastRuns on your balance at the prices above. New accounts start with free usage.
Settings
Cohere: Embed 5 Fast is a multimodal embedding model from Cohere, positioned as the lower-latency, cost-efficient tier of the company's Embed 5 family. It is built for developers who need to turn content into vector representations at scale, where throughput and responsiveness matter as much as embedding quality. The model sits in the embeddings category rather than the chat category, so it does not generate conversational text; its job is to produce vectors that downstream systems can search, compare or classify.
The model accepts text, images, and interleaved text and image content as input, and returns embeddings as its only output. That multimodal input support means a single model can embed a paragraph, a product photo, or a document page that mixes both, which keeps retrieval pipelines simpler than maintaining separate text and image encoders. It offers a context length of 128,000 tokens, with a maximum output of 115,200, giving it room to handle long documents and large image-and-text payloads without aggressive chunking. At roughly 1B parameters, it is a compact model by design, which is consistent with its fast, cost-efficient positioning.
Those capabilities point to a fairly specific set of use cases. Semantic search over mixed media is the most direct fit: indexing text and images into one vector space so a query can retrieve either. Retrieval-augmented generation pipelines can use it to embed knowledge bases that contain figures, screenshots or scanned pages alongside prose. Recommendation and similarity systems benefit from the shared space, since items can be compared regardless of whether their content is textual or visual. Classification, clustering, deduplication and anomaly detection over large corpora are also natural applications, particularly where the volume of data makes latency and cost per embedding a practical concern.
You call the model through api14's OpenAI-compatible API using the model id "cohere/embed-5-fast". Because the endpoint follows the OpenAI embeddings convention, existing clients and SDKs that already target that interface can point at api14 with minimal changes, typically just swapping the base URL and the model id. The model does not declare any supported request parameters beyond the standard embedding request shape, so you should not expect tuning knobs such as custom dimensions or provider-specific options to be available; send your input and read the returned vectors.
Given that it is the fast tier of the Embed 5 family, it is the sensible default when you are embedding at volume and latency or throughput is the constraint. If your workload is purely text and does not need image handling, a text-only embedding model may be a simpler fit. If you need a model that generates text rather than vectors, this is not the right choice, since its output modality is embeddings only.
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-fast",
"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 Fast.