Embed 5 Fast
Embedding modelCard summarizedA faster, lighter version of embed-v5.0-pro
Source: Embed 5 launch post and Embed models docs (no model card published)
Where to call it
- Cohere API
embed-v5.0-fast - Microsoft Foundry / Azure
—
Good for
- Live query paths: interactive search and agent loops
- High-volume RAG where latency and cost matter
- Querying a corpus that was indexed with Embed 5 Pro
Overview
Cohere · 01Fast, front
Embed 5 Fast
A faster, lighter version of embed-v5.0-pro
Typical price
$0.08 / 1M tok
Text input
- Dimensions
- 2048
- Max input
- 128K tokens
- Modes
- text embeddings, image embeddings
- Released
- 30 Sep 2026
Embed 5 Fast is one of two tiers in Cohere's Embed 5 family. Fast is the latency and cost tier, aimed at interactive search, high-volume RAG and agent loops; Cohere measures it at 2.4x Pro's throughput. Both tiers embed text, images and fused text-image inputs into one shared space, so a corpus indexed with Pro can be queried with Fast. They support 100+ languages, a 128K-token context, Matryoshka dimensions from 256 to 2048, and float, int8 and binary outputs. launch post; models docs
embed-v5.0-fast- Typical price
- $0.08 / 1M tok
- Dimensions
- 2048 (default; can be shortened to 1536, 1024, 768, 512 or 256)
- Max input
- 128K tokens
- Languages
- 100+ languages
- Weights
- Closed
- Released
- 30 Sep 2026
- text embeddings
- image embeddings
- multimodal embeddings (fused text + image, e.g. PDF pages)