Cohere Embed English v3
1024-dim managed embedding with task-specific input types
About
Cohere's embed-english-v3.0 is the managed embedding model that defined the company's retrieval stack before Embed v4: fixed 1024-dimensional vectors, a 512-token context window, and a required input_type parameter (search_document, search_query, classification, clustering) that tunes the representation per task. It also accepts images as base64 input for cross-modal use. Available on Cohere's API, AWS Bedrock, and Azure, it was long the go-to for teams wanting a vendor-managed alternative to OpenAI with explicit retrieval semantics. It is showing its age now: the 512-token window forces heavy chunking, dimensions are fixed (no Matryoshka), and Cohere itself has moved the lineup to Embed v4 — but it remains widely deployed and well documented.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
MTEB average ~64.5 (reported in third-party comparisons)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: 1024 fixed dims, 512-token window, required input_type parameter, and image-input support verified in Azure deployment docs and Bedrock model documentation read this session.; Pricing ($0.10/1M) is third-party reported — labeled as such.; Superseded by Embed v4 — disclosed honestly in the entry..