Turbopuffer vs Voyage AI

A side-by-side look at pricing, capabilities, pros, cons, and our editorial scores.

	Turbopuffer RAG	Voyage AI RAG
Tagline	Fast search on object storage	State-of-the-art embedding models and rerankers purpose-built for retrieval and RAG.
Category	RAG	RAG
Pricing	Paid· launch: $16/month · scale: $256/month · enterprise: >=$4,096/month	Freemium· Free tier: 200M free text tokens per account for current models (50M for older specialized). Text embeddings $0.00002–$0.00018 per 1K tokens depending on model tier. Rerankers $0.00002–$0.00005 per 1K tokens after 200M free. Multimodal $0.12 per 1M text tokens + $0.60 per 1B pixels. Batch API 33% discount. File storage $0.05/GB/month.
Model	bring-your-own embeddings (any provider)	in-house (voyage-3.5, voyage-4 series, voyage-code-3, voyage-finance-2, voyage-law-2, voyage-multimodal-3.5, voyage-context-3, rerank-2.5)
Editorial score	—	—
Use cases	Production RAG chatbotsMulti-tenant semantic searchAgent long-term memorySemantic code searchRecommendation systemsLog and observability searchHybrid keyword + vector product searchLarge-scale document retrievalRe-embedding experiments via namespace branching	Production RAG chatbot over proprietary docsTwo-stage retrieval with embed + rerankCode search across a monorepoLegal contract semantic searchFinancial filings and research retrievalMultimodal image-and-text searchLong-context document embedding (32K tokens)Context-aware chunk embedding for dense passagesBatch embedding of large historical corporaMongoDB Atlas Vector Search backends
Pros	Object-storage-first architecture is dramatically cheaper than RAM-resident vector DBs at billion-vector scale Native hybrid search (vector + BM25) with metadata filters in a single query Namespace model plus copy-on-write branching maps cleanly to multi-tenant RAG and re-embedding workflows Very high write and query throughput demonstrated in production (10M+ writes/s, 25k+ QPS) Serverless — no clusters, shards, or replicas to manage; scales namespaces automatically Used in production by demanding AI teams (Cursor, Notion, Anthropic, Linear), which is meaningful social proof Comprehensive REST API and clear latency/recall SLIs published rather than hand-waved	Consistently near the top of MTEB and BEIR retrieval leaderboards — measurable recall gains over OpenAI text-embedding-3-large in most public evaluations. Short output dimensions (as low as 256 or 512) cut vector storage and ANN latency 3x–8x versus 1536/3072-dim competitors. Domain-tuned models (code, finance, legal) meaningfully outperform general embeddings on in-domain corpora. voyage-context-3 embeds chunks with awareness of surrounding document context, reducing the classic 'lost context' problem in fixed-window chunking. Rerank-2.5 with instruction-following gives a clean two-stage retrieval pipeline without training a custom cross-encoder. Generous 200M-token free tier per account makes prototyping and small production workloads essentially free. Batch API offers a 33% discount for large offline embedding jobs. MongoDB acquisition (2025) means tight, ongoing integration with Atlas Vector Search.
Cons	No free tier and a $16/mo floor even on the smallest plan, so it is not a fit for hobby projects or evaluation on a shoestring Closed-source, managed-only — no self-host option, which rules out air-gapped or fully sovereign deployments below the Enterprise BYOC tier Object-storage cold reads mean tail latency and cache-miss behaviour matter more than in a purely in-memory system; tuning matters for latency-critical UX You bring your own embeddings — no built-in embedding model, ingestion pipeline, or chunking, unlike higher-level RAG platforms Enterprise features people often need in regulated industries (SSO, HIPAA BAA, audit logs) start at the $256/mo Scale plan and above	API-only closed models — no self-hosting option, so latency-sensitive or air-gapped deployments are ruled out. Not an end-to-end RAG stack — you still need a vector database, LLM, and orchestration layer, which increases integration surface. Post-MongoDB acquisition, product roadmap and standalone longevity depend on MongoDB's priorities. Domain models cover finance, legal, and code but nothing else — medical, scientific, or multilingual-heavy corpora fall back to general models. Documentation is competent but sparser than OpenAI's or Cohere's — fewer end-to-end recipes for advanced patterns like hybrid search or query expansion. Pricing per token is competitive but not the cheapest — self-hosted open models (e.g. BGE, E5) are free at inference if you have GPUs.
Website	turbopuffer.com	www.voyageai.com

Pick Turbopuffer if

✅ Object-storage-first architecture is dramatically cheaper than RAM-resident vector DBs at billion-vector scale
✅ Native hybrid search (vector + BM25) with metadata filters in a single query
✅ Namespace model plus copy-on-write branching maps cleanly to multi-tenant RAG and re-embedding workflows
✅ Very high write and query throughput demonstrated in production (10M+ writes/s, 25k+ QPS)

Pick Voyage AI if

✅ Consistently near the top of MTEB and BEIR retrieval leaderboards — measurable recall gains over OpenAI text-embedding-3-large in most public evaluations.
✅ Short output dimensions (as low as 256 or 512) cut vector storage and ANN latency 3x–8x versus 1536/3072-dim competitors.
✅ Domain-tuned models (code, finance, legal) meaningfully outperform general embeddings on in-domain corpora.
✅ voyage-context-3 embeds chunks with awareness of surrounding document context, reducing the classic 'lost context' problem in fixed-window chunking.

Compare a different pair →