Cohere Embed v4 Pricing
Cohere Embed v4 is an embedding model from Cohere, last listed at $0.1000 per 1M tokens. It produces 1024-dimensional vectors and accepts inputs up to 128,000 tokens.
Status checked against official pricing pages. Sources: Cohere
No longer on the official price list. Prices shown are the last ones published. Cohere’s pricing page no longer publishes a per-token price for Embed v4; its docs list Embed v5 as the newer model. Newer option: Cohere Embed v5.
What does it cost to embed a corpus with Cohere Embed v4?
Embedding cost is a one-time charge per document — unlike LLM inference, you only pay to embed each chunk once at ingest time. Re-embedding only happens when you re-chunk or swap models.
| Corpus size | Tokens | One-time embed cost |
|---|---|---|
| Small docs corpus (~1M tokens) | 1,000,000 | $0.1000 |
| Mid-size corpus (~50M tokens) | 50,000,000 | $5.0000 |
| Enterprise corpus (~500M tokens) | 500,000,000 | $50.0000 |
Want to model your RAG stack end-to-end? RAG cost calculator →
Vector database sizing — what 1024-dim vectors cost to store
Each Cohere Embed v4 embedding is a 1024-dimensional float32 vector, ~ 4.00 KB per vector. At 1M vectors that's ~ 3906.3 MB of raw vector data (before index overhead). Compare against alternates with different dimension counts:
- OpenAI text-embedding-3-small: 1536 dims → 5859.4 MB per 1M vectors
- OpenAI text-embedding-3-large: 3072 dims → 11718.8 MB per 1M vectors
- Voyage 3 Large: 2048 dims → 7812.5 MB per 1M vectors
- Voyage 3 Lite: 512 dims → 1953.1 MB per 1M vectors
- Cohere Embed v4: 1024 dims → 3906.3 MB per 1M vectors
- Gemini Embedding: 768 dims → 2929.7 MB per 1M vectors
Alternatives to Cohere Embed v4
Cheaper embedding models
Higher-priced embedding models
FAQ
How much does Cohere Embed v4 cost?
Cohere no longer publishes a price for Cohere Embed v4. No longer on the official price list. Prices shown are the last ones published. Cohere’s pricing page no longer publishes a per-token price for Embed v4; its docs list Embed v5 as the newer model. Newer option: Cohere Embed v5.
What are Cohere Embed v4's embedding dimensions?
Cohere Embed v4 outputs 1024-dimensional vectors. This affects both storage size and similarity computation speed in your vector database.
What is the max input length?
Cohere Embed v4 accepts inputs up to 128,000 tokens. Larger documents must be chunked before embedding.
What's the cheapest alternative?
The cheapest comparable model is OpenAI text-embedding-3-small at $0.0200 per 1M tokens.