Cohere · Embedding
Embed V5.0 Fast
Cohere's low-latency embedding model for interactive search and high-volume retrieval. Embed 5 Fast and Pro share a vector space: index with Pro and query with Fast using the same dimensions. Supports text, images, mixed text/image items, over 100 languages, and a 128,000-token context. Choose 256, 512, 768, 1024, 1536, or 2048 dimensions (default), with float or base64 output. Set input_type to search_query, search_document, classification, clustering, or image. Send up to 96 items per request. Images use PNG, JPEG, or WebP URLs or data URLs, with up to 16 images, 8 MiB per image and 16 MiB combined; images are resized to fit 1,120 pixels. Text and image tokens are billed separately. Image requests require account or team balance; x402 is not supported.
Evidence and sources
Released 30 Sep 2026Maker HQ: US (according to Cohere's own website)
All rankings:
#354 on OpenRouter this week
#415 on OpenRouter this month
#427 on OpenRouter todaySources: Cohere API data, LiteLLM data, Models.dev data, NanoGPT, llminfo researchPrimary data from Cohere API
from LiteLLM: deprecation date
from Models.dev: weights availability
from NanoGPT: description
from llminfo research: release date
Advantages and trade-offs
Compared with 10 representative embedding models
Advantages
Trade-offs
Deprecation scheduled
The maker plans to deprecate it on 23 Jan 2027, within the next 180 days.
No open weights
The weights aren't public, so you can't run it yourself.
Input & output
as stated by the model's providersText
Images
Embeddings
Also known as: azure_ai/cohere-embed-v5-fast, cohere-embed-v5-fast, cohere/embed-5-fast, cohere/embed-v5.0-fast, embed-5-fast
Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works