Skip to content

Voyage 4 Large

Voyage 4 Large is Voyage AI's Voyage 4 flagship embedding model. It uses a mixture-of-experts (MoE) architecture. Voyage AI reports state-of-the-art general retrieval in their published benchmarks, with serving costs about 40% lower than comparable dense models, and average gains over OpenAI text-embedding-3-large, Cohere Embed v4, and Gemini Embedding 001 in the same comparison. It shares one embedding space with voyage-4 and voyage-4-lite.

index.ts
import { embed } from 'ai';


const result = await embed({
  model: 'voyage/voyage-4-large',
  value: 'Sunny day at the beach',
})

What To Consider When Choosing a Provider

  • Configuration: Voyage 4 Large targets teams that need top published scores and can pay for the flagship on the paths that matter (often document embedding).
  • Configuration: Pair Voyage 4 Large document vectors with smaller Voyage 4 query models when query volume is high.
  • Configuration: Treat a move to Voyage 4 as a new index. Test on a sample corpus before you re-embed everything.
  • Zero Data Retention: AI Gateway does not currently support Zero Data Retention for this model. See the documentation for models that support ZDR.
  • Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.

When to Use Voyage 4 Large

Best For

  • Corpus embedding once: You want maximum document-side quality in Voyage AI's published Voyage 4 results
  • Asymmetric RAG: Documents use Voyage 4 Large and queries use voyage-4-lite
  • Enterprise search: Long documents within the window of 32K tokens
  • Upgrades from voyage-3-large: You accept a full re-embed for Voyage 4's shared space and MoE gains

Consider Alternatives When

  • Lower per-query cost: Use voyage-4 or voyage-4-lite for queries, or symmetric indexing with a smaller Voyage 4 model
  • Mid-tier symmetric use: voyage-4 when you want one model for both sides
  • Code-only corpora: Use voyage-code-3 for repositories where source code is the primary content type
  • Multimodal embeddings: Pick a model with native image inputs

Conclusion

MoE architecture gives Voyage 4 Large flagship retrieval accuracy at lower serving costs than dense alternatives. Use it for document embeddings and pair with lighter Voyage 4 models on queries to optimize per-request spend through AI Gateway.

Frequently Asked Questions

  • What is the difference between Voyage 4 Large and voyage-4?

    Voyage 4 Large is the MoE flagship with the highest average scores in Voyage AI's published Voyage 4 comparison. voyage-4 is the mid-sized model. Both share the same embedding space as voyage-4-lite.

  • How does Voyage 4 Large compare to voyage-3-large?

    Voyage AI reports better retrieval accuracy than voyage-3-large at a lower price, using MoE and the Voyage 4 training stack. Moving from Voyage 3 to Voyage 4 requires re-embedding because the embedding space changes.

  • What is the context window for Voyage 4 Large?

    32K tokens. Size chunks so single requests stay under this limit on long documents.

  • When should I use Voyage 4 Large over voyage-4-lite?

    Use Voyage 4 Large when you need the strongest published Voyage 4 vectors, especially for one-time or infrequent document embedding. Use voyage-4-lite when you want fewer parameters for queries or symmetric indexing at lower compute.

  • How do I access Voyage 4 Large through Vercel AI Gateway?

    Add your Voyage AI API key in AI Gateway settings, then send embedding requests through AI Gateway. AI Gateway authenticates requests and records usage.

  • Do I need to re-embed my data to switch from voyage-3-large?

    Yes. Moving from Voyage 3 to Voyage 4 requires re-embedding because the embedding space is new. Within Voyage 4, you can often keep voyage-4-large document vectors and change query models if you use asymmetric retrieval.

  • Is Voyage 4 Large suitable for RAG applications?

    Yes. Voyage AI positions it for retrieval-augmented generation and high-accuracy document indexing, including asymmetric setups where queries use a smaller Voyage 4 model.

  • What is mixture-of-experts in Voyage 4 Large?

    Voyage 4 Large routes tokens through expert subnetworks so Voyage AI can raise accuracy while reporting serving costs about 40% lower than comparable dense models.