Voyage AI Embedding Pricing: voyage-4, voyage-4-large & voyage-4-lite (June 2026)
The voyage-4 family (January 2026) is Voyage's current generation and tops its own RTEB retrieval benchmark. Here is what it costs, when each tier is worth it, and the shared-vector-space feature that lets you mix models without re-indexing.
How much does Voyage AI cost, and is there a free tier?
Voyage AI embeddings cost $0.02 per million tokens for voyage-4-lite, $0.06/M for voyage-4, and $0.12/M for voyage-4-large. Every account gets 200 million free tokens on the voyage-4 generation (about 400,000 documents at zero cost), and the Batch API takes 33% off every tier.
All voyage-4 models share one vector space, so you can index with voyage-4-large and query with voyage-4-lite without re-indexing. The legacy voyage-3.x models no longer include the free allocation.
Current Pricing
| Model | $/M tokens | Batch $/M | Dims | Context | Free |
|---|---|---|---|---|---|
voyage-4-large Flagship (MoE) | $0.12 | $0.080 | 1,024 | 32,000 | 200M |
voyage-4 Recommended | $0.06 | $0.040 | 1,024 | 32,000 | 200M |
voyage-4-lite Cost optimized | $0.02 | $0.013 | 1,024 | 32,000 | 200M |
voyage-code-3 Code search | $0.18 | $0.120 | 1,024 | 32,000 | 200M |
voyage-context-3 Contextual chunks | $0.18 | $0.120 | 1,024 | 32,000 | 200M |
voyage-law-2 Legal RAG | $0.12 | $0.080 | 1,024 | 16,000 | 50M |
voyage-finance-2 Financial RAG | $0.12 | $0.080 | 1,024 | 16,000 | 50M |
voyage-4 models default to 1024 dimensions with Matryoshka options (256/512/1024/2048). Batch jobs complete within 12 hours (33% off standard rate). Free column is the per-account lifetime allocation.
Why voyage-4 is the Price-Accuracy Sweet Spot
At $0.06/M tokens, voyage-4 costs 3x more than OpenAI text-embedding-3-small but approaches the retrieval quality of the previous flagship voyage-3-large at a mid-sized model's efficiency. For RAG applications where retrieval quality directly affects answer quality, that is a meaningful step up for English and multilingual content alike.
The top tier, voyage-4-large ($0.12/M), is the first production-grade embedding model to use a mixture-of-experts (MoE) architecture. On Voyage's RTEB benchmark (29 retrieval datasets across 8 domains) it outperforms OpenAI text-embedding-3-large by 14% and Cohere embed-v4 by 8.2% on NDCG@10, while MoE keeps serving costs roughly 40% below comparable dense models.
Use voyage-4 batch at $0.04/M for document indexing, and keep standard ($0.06/M) for real-time query embedding. At 1 billion indexing tokens + 10M monthly query tokens, your total monthly cost is $40 indexing (one-time) + $0.60 query = under $1/month ongoing after initial indexing.
The Shared-Vector-Space Feature
Built into voyage-4: All voyage-4 models (lite, 4, large) embed into the same vector space. You can index documents with voyage-4-large and query with voyage-4-lite without re-indexing.
Practical application: index your corpus once with voyage-4-large ($0.12/M, maximum quality) and serve real-time queries with voyage-4-lite ($0.02/M, fast). You pay the premium once for indexing, then query cheaply forever. For a 1B token corpus indexed once + 100M monthly query tokens: one-time index cost = $120, monthly query cost = $2.
Domain-Specific Models: When the Premium Pays
Code search, documentation RAG, GitHub issue retrieval. Trained on code repositories.
Legal document search, contract review. Trained on legal corpora with jurisdiction awareness.
Financial document retrieval, earnings call search, SEC filing analysis.
Free Tier: 200M Lifetime Tokens on voyage-4
Every Voyage AI account receives 200 million free tokens on the voyage-4 generation (voyage-4-large, voyage-4, voyage-4-lite) plus voyage-context-3 and voyage-code-3, and 50 million free tokens on the voyage-2 domain models. At 500 tokens per document, 200M tokens is 400,000 documents you can embed at zero cost - enough to build and test an entire prototype RAG pipeline before spending anything. No card required. The older voyage-3.x models no longer include a free allocation.
Previous Generation: voyage-3.x
The voyage-3 series remains callable for existing indexes, with prices unchanged, but Voyage states the voyage-4 models are strictly better in quality, context length, latency, and throughput. The voyage-3.x models no longer include the 200M free allocation.
| Legacy model | $/M tokens | Dims | MTEB |
|---|---|---|---|
| voyage-3-large | $0.18 | 2,048 | 68.9 |
| voyage-3.5 | $0.06 | 1,024 | 67.1 |
| voyage-3-lite | $0.02 | 512 | 61.7 |