Language
Embeddings
Turn text or images into vectors for search, RAG and clustering.
What it is
An embedding model converts a piece of text (or an image) into a list of numbers, a vector, so that similar meanings land close together. It powers semantic search, retrieval-augmented generation (RAG), recommendations, deduplication and clustering.
How pricing works
Billed per input token; there is no output text. Prices are typically 10–100× lower than LLMs. Storage of the resulting vectors is a separate cost (see Vector databases).
We compare everything in one unit: USD per 1M input tokens.
Typical uses
- Semantic search over documents
- RAG: finding the right context to feed an LLM
- Recommendations and similarity
- Clustering and deduplication
What to compare
- Retrieval quality (e.g. MTEB benchmark)
- Vector dimensions (storage cost)
- Max input length
- Multilingual support
Watch out for
- Switching embedding models means re-embedding your whole corpus.
- Higher dimensions cost more storage and search time.
All 7 APIs, cheapest first
Free to browseTick 2 to 4 APIs to compare them side by side.
| API | $ / 1M tokens | Dimensions | Max input | Multilingual | Quality | Price data | Compare |
|---|---|---|---|---|---|---|---|
| text-embedding-3-small OpenAI | $0.02 | 1,536 | 8,191 tokens | Yes | 70 | Verified Oct 4, 2026 · aging | |
| Titan Text Embeddings V2 Amazon Web Services | $0.02 | 1,024 | 8,192 tokens | Yes | – | Unverified sample | |
| jina-embeddings-v4 Jina AI | $0.05 | 2,048 | 32K tokens | Yes | 80 | Unverified sample | |
| voyage-4 Voyage AI | $0.06 | 1,024 | 32K tokens | Yes | 85 | Verified Oct 4, 2026 | |
| Qwen text-embedding-v4 Alibaba Cloud (Qwen) | $0.07 | 1,024 | 8,192 tokens | Yes | – | Unverified sample | |
| Embed v4 Cohere | $0.12 | 1,536 | 128K tokens | Yes | 84 | Unverified sample | |
| Gemini Embedding Google (Gemini / Vertex) | $0.15 | 3,072 | 2,048 tokens | Yes | – | Unverified sample |