Voyage 3 Large
The commercial retrieval specialist — top MTEB with quantization-aware Matryoshka embeddings.
Acquired — still operating
acquired by MongoDB 2025
About
Voyage-3-Large (late 2024, Voyage AI) is the commercial retrieval specialist: top-tier MTEB performance with Matryoshka embeddings (truncate dimensions without re-encoding) and quantization-aware training for efficient serving. Voyage was acquired by MongoDB (2025) — the models now ship inside MongoDB's Atlas vector search story.
Closed and commercial: API-only, no weights. Pricing is per-token; the MongoDB acquisition suggests deepening Atlas integration. For teams that want best-in-class retrieval without managing embedding infrastructure, Voyage 3 Large (or the newer Voyage 4) is a top commercial answer alongside Cohere and OpenAI. The acquisition is a stability signal but also a roadmap question — verify current positioning.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-claimed: top MTEB among commercial embeddings — treat as provider-claimed. Independent: strong third-party reception for retrieval quality.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier). Top commercial retrieval quality with Matryoshka flexibility, and the MongoDB acquisition is a stability signal. Docked three points: (1) closed API-only — no weights; (2) benchmark claims are provider-reported; (3) the acquisition creates roadmap uncertainty — verify current model positioning. For buy-don't-build retrieval.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 4.5/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · provider-tier estimate — thinner evidence (no public adoption metric)
- · top commercial retrieval reputation; MongoDB acquisition (2025) is a stability signal
- · released 700 days ago
- · manual review: Matryoshka embeddings are genuinely useful; acquisition creates roadmap questions; license: proprietary — API only, no weights