Kuraberu AI models, home

AI models

Sections

Embedding model prices

14 models from 30 providers · Prices checked

The cheapest embedding models

every model

Per million input tokens, at the cheapest provider a ranking may quote: never a free tier, never a price nobody accounts for.

  1. 1

    text-embedding-3-small

    OpenAI · 8k context · closed weights · 9 providers

    $0.02/M tokens
  2. 2

    Embed v1 0.6b

    Perplexity · 32k context · closed weights · 2 providers

    0.4¢/M tokens
  3. 3

    Qwen3 Embedding 4B

    Hugging Face Inference Providers · 33k context · open weights · 6 providers

    $0.01/M tokens
  4. 4

    BGE M3

    DigitalOcean Gradient · 8k context · open weights · 1 provider

    $0.02/M tokens
  5. 5

    Qwen3 Embedding 8B

    TrustedRouter · 33k context · open weights · 14 providers

    $0.02/M tokens
  6. 6

    Embed v1 4b

    Perplexity · 32k context · closed weights · 2 providers

    $0.03/M tokens
  7. 7

    Qwen3 Embedding 0.6B

    DigitalOcean Gradient · 33k context · open weights · 6 providers

    $0.04/M tokens
  8. 8

    BGE Multilingual Gemma2

    Infomaniak · 8k context · closed weights · 2 providers

    a new price, being checked
    $0.08/M tokens
  9. 9

    Mistral Embed

    Mistral AI · 8k context · closed weights · 3 providers

    $0.10/M tokens
  10. 10

    text-embedding-ada-002

    OpenAI · 8k context · closed weights · 7 providers

    $0.10/M tokens

What one price buys

Priced per 1M tokens. One million tokens of text turned into vectors, standard tier, no batch discount. Prices are per million input tokens in US dollars; an embedding model writes no output tokens.

Models

14

with a page

Prices compared

73

side by side

Providers

30

selling them

Typical price

$0.10

per 1M tokens

Providers selling embedding models

all providers

Every embedding model

Qwen3 Embedding 8Btext-embedding-3-largetext-embedding-3-smallGemini Embedding 001text-embedding-ada-002Gemini Embedding 2Qwen3 Embedding 0.6BQwen3 Embedding 4BMistral EmbedBGE Multilingual Gemma2Embed v4.0Embed v1 0.6bEmbed v1 4bBGE M3

How we count

An embedding model turns text into vectors and writes no tokens back, so its price is the input price alone, per million tokens, and never the three-to-one blend the text models are compared on. The two are never ranked against each other.

Every price is measured against one list price per model: the lab's own wherever the lab sells it, otherwise an official cloud's or OpenRouter's. A price below that with nothing published to account for it is shown and not ranked. An open-weight model the lab never hosted keeps its page on two providers' rows alone, because the market of hosts is the finding. Rerankers are not here yet: their makers price them per query and a few hosts per token, and one page cannot hold both. The method