Amazon Titan Text Embeddings V2

Embedding modelCard summarized

Second-generation text embeddings model with configurable output dimensions and improved accuracy for retrieval tasks

Source: AWS AI Service Card: Amazon Titan Text Embeddings (G1 and V2, current as of 3 Dec 2024)

Use it via API

$0.02 / 1M tok

Code example Compare

Where to call it

Amazon Bedrockamazon.titan-embed-text-v2:0

Supports

Batch API

Full feature support

Good for

Question answering with RAG
Search optimization (semantic matching alongside keyword search)
Content grouping by clustering

Overview

Amazon · 01Embeddings

Titan Text V2

Second-generation text embeddings model with configurable output dimensions and improved accuracy for retrieval tasks

Typical price

$0.02 / 1M tok

Text input

Dimensions: 1024
Max input: 8,192 tokens
Modes: text embeddings, code retrieval
Released: 30 Apr 2024

Titan Text Embeddings V2 is Amazon's second-generation text embedding model on Amazon Bedrock, optimized for RAG, search and content grouping. It accepts up to 8,192 tokens, returns smaller vectors than G1 with optional binary output, and is trained on 112 natural and 13 programming languages. AWS reports a small MTEB gain over G1 (0.60 vs 0.59 overall, 0.51 vs 0.47 retrieval) and lower measured stereotype bias. service card

amazon.titan-embed-text-v2:0

Typical price: $0.02 / 1M tok
Dimensions: 1024 (default), 512 or 256
Max input: 8,192 tokens
Languages: Trained on 112 languages (testing most extensive in English)
Weights: Closed
Released: 30 Apr 2024

text embeddings
code retrieval