Voyage-3 is a family of general-purpose text embedding models developed by Voyage AI, launched in September 2024 with voyage-3 and voyage-3-lite [1], expanded in January 2025 with voyage-3-large [2], and refreshed in May 2025 with voyage-3.5 and voyage-3.5-lite [3]. All members share a dense-vector retrieval interface, a 32,000-token context window, and a 1024-dimensional default output, while differing in cost, peak quality, and the dimensions to which their vectors can be truncated. At its January 2025 launch, voyage-3-large led the public Massive Text Embedding Benchmark (MTEB) English leaderboard with an average score of 65.1 [2], and Voyage AI reports the family outperforms OpenAI's text-embedding-3-large by 7.55 to 9.74 percent on its own retrieval evaluations across legal, financial, code, technical-documentation, multilingual, and long-document domains [1][2].

Voyage-3 was the first generation in which Voyage AI shipped Matryoshka representation learning and quantization-aware training as default features [2]. With voyage-3-large and voyage-3.5, a single API call can return embeddings of 256, 512, 1024, or 2048 dimensions in float32, int8, uint8, or binary precision without re-encoding [2][3]. The family is closely associated with Anthropic, which since 2024 has documented Voyage AI as its recommended embedding provider for Claude-based retrieval-augmented generation and has stated that it does not train its own embedding model [9]. Voyage AI was not acquired by Anthropic: the company was bought by MongoDB for 220 million dollars in a cash-and-stock deal announced on 24 February 2025 [7][20], and now operates as Voyage AI by MongoDB. As of 2026, the voyage-3 line has been succeeded by the voyage-4 family and is classified in Voyage AI's pricing as an older generation [6][18].

Background and naming

Voyage AI was founded in September 2023 by Stanford computer-science professor Tengyu Ma with co-founders Hong Liu and Kaidi Cao [21]. The company's early product was the voyage-01 and voyage-lite-01 embedding models released in late 2023, followed by domain-specialized checkpoints such as voyage-law-2 (April 2024) and voyage-finance-2 (May 2024). The voyage-2 line was a moderate iteration on the original architecture, but it was the voyage-3 release in September 2024 that established the company as a peer of OpenAI and Cohere in independent benchmarks [1]. By early 2025 the company's models were used by more than 250 customers, and MongoDB's acquisition announcement named Anthropic, LangChain, Harvey, and Replit among them [7][21].

The "3" in the name refers to the third major generation of Voyage's general-purpose family rather than to any specific architecture, parameter count, or training recipe. The naming convention places quality tiers after the version number: the base name (voyage-3, voyage-3.5) denotes the standard quality model, the suffix -lite denotes a smaller checkpoint trained partly via distillation, and the suffix -large denotes the highest-quality flagship. The minor version step from 3 to 3.5 reflects a refreshed training mixture and distillation from voyage-3-large rather than a new core architecture [3].

Model line

voyage-3 and voyage-3-lite (September 2024)

Voyage AI announced voyage-3 and voyage-3-lite on 18 September 2024 [1], positioning them as drop-in replacements for OpenAI's text-embedding-3-large with much smaller default embedding dimensions and a four-times-longer context window. Voyage AI reported that voyage-3 outperformed text-embedding-3-large by an average of 7.55 percent on a suite of 40 retrieval datasets covering eight domains, while costing 2.2 times less per token and producing vectors three times smaller than the OpenAI default of 3072 dimensions [1]. The lite variant was reported as 3.82 percent more accurate than text-embedding-3-large on the same datasets while costing 6.5 times less, with vectors six times smaller [1]. Voyage AI stated that "voyage-3 outperforms OpenAI v3 large across all eight evaluated domains" [1].

Both models were trained with an improved transformer encoder architecture, distillation from larger teacher checkpoints, more than two trillion tokens of pre-training data, and retrieval-result alignment via human feedback [1]. Both accept up to 32,000 tokens of input. The 1024-dimensional output of voyage-3 and the 512-dimensional output of voyage-3-lite were fixed at launch; Matryoshka truncation arrived with the later voyage-3-large model [1][2].

voyage-3-large (January 2025)

Voyage-3-large launched on 7 January 2025 as the new state-of-the-art flagship in the family [2]. It was the first Voyage model to ship Matryoshka and quantization-aware training as default features, allowing a single response at 256, 512, 1024, or 2048 dimensions in float32, int8, uint8, binary, or unsigned-binary precision [2]. Voyage AI described it as "the new state-of-the-art general-purpose and multilingual embedding model" and reported that it outperformed OpenAI text-embedding-3-large by 9.74 percent on average and Cohere Embed v3 English by 20.71 percent across 100 datasets in eight domains, and led the public MTEB leaderboard with an overall score of 65.1, edging out NV-Embed v2 by 0.3 points and OpenAI text-embedding-3-large by 0.5 points [2].

The model remained competitive after aggressive compression. At 1024 dimensions in int8 it scored only 0.31 percent below full-precision float at 2048 dimensions while using eight times less storage, and its binary embeddings at 512 dimensions stayed 1.16 percent above OpenAI text-embedding-3-large (float32, 3072 dimensions) while requiring 200 times less storage, which Voyage AI illustrated as reducing a $20,000 monthly vector-database bill to about $100 [2]. Voyage-3-large was also reported to surpass voyage-law-2 on legal retrieval and voyage-finance-2 on financial retrieval, although the specialized voyage-code-3 remained the preferred model for source-code retrieval [2][4].

voyage-3.5 and voyage-3.5-lite (May 2025)

Voyage AI released voyage-3.5 and voyage-3.5-lite on 20 May 2025 [3]. The 3.5 step did not introduce a new flagship; instead it lifted the quality of the mid-tier and entry-tier models while leaving their per-token prices identical to voyage-3 and voyage-3-lite. Voyage AI reported that voyage-3.5 was 8.26 percent more accurate than OpenAI text-embedding-3-large, 2.66 percent more accurate than voyage-3, and 1.63 percent more accurate than Cohere Embed v4 on average across the same 100 datasets used to evaluate voyage-3-large [3]. The lite variant was reported as 6.34 percent better than OpenAI text-embedding-3-large and 4.28 percent better than voyage-3-lite, reaching quality "within 0.3% of Cohere-v4 at 1/6 the cost" [3].

Both voyage-3.5 models adopted the same Matryoshka and quantization toolkit introduced with voyage-3-large, supporting 256, 512, 1024, and 2048 dimensions at float32, int8, uint8, and binary precision [3]. Voyage AI attributed the quality gains to "an improved mixture of training data, distillation from voyage-3-large, and the use of Voyage AI rerankers" [3]. The company calculated that a 2048-dimensional int8 voyage-3.5 embedding cost roughly 83 percent less to store and search than a 3072-dimensional float32 OpenAI text-embedding-3-large embedding, while still achieving higher retrieval quality [3].

Models, dimensions, and pricing

All models in the voyage-3 family accept up to 32,000 tokens of input per request and share a single REST endpoint at api.voyageai.com [5]. The default output dimension of every model is 1024; only voyage-3-large, voyage-3.5, and voyage-3.5-lite expose additional Matryoshka dimensions and quantized precisions [2][3]. Voyage-3-lite predates the Matryoshka toolkit and is a fixed 512-dimensional float32 model [1]. Pricing below is for the public API as listed on Voyage AI's documentation [6]. When the family launched, the first 200 million tokens per account were free; as of 2026 that free allowance applies to the current voyage-4 generation, and Voyage AI's pricing page classifies the voyage-3 models as an older generation without free tokens [6].

Model	Released	Default dim	Other dims (Matryoshka)	Context	Price per 1M tokens
voyage-3-lite	Sep 2024	512	fixed	32,000	$0.02
voyage-3	Sep 2024	1024	fixed	32,000	$0.06
voyage-3-large	Jan 2025	1024	256, 512, 2048	32,000	$0.18
voyage-3.5-lite	May 2025	1024	256, 512, 2048	32,000	$0.02
voyage-3.5	May 2025	1024	256, 512, 2048	32,000	$0.06

With Matryoshka truncation enabled, the same call can yield, for example, a 256-dimensional int8 vector that occupies 256 bytes instead of the 4096 bytes of a 1024-dimensional float32 vector, a 16-fold reduction in vector-database storage [14]. Binary embeddings, in which each dimension is reduced to a single bit, compress further to roughly one two-hundredth of the float32 baseline [2]. Voyage AI's reported retrieval scores for these compressed embeddings remain competitive with full-precision OpenAI baselines, although accuracy still drops modestly relative to full-precision Voyage embeddings of the same dimension [2].

Matryoshka representation learning

Matryoshka representation learning, introduced by Aditya Kusupati and colleagues in 2022, trains a single embedding model with a sum of contrastive losses computed over nested prefixes of the output vector [12]. The result is an embedding in which the first k coordinates are themselves a self-contained representation for many values of k, so the vector can be truncated to a smaller dimension without retraining and with only a modest loss in retrieval quality. OpenAI's text-embedding-3 series, released in January 2024, was the first widely deployed commercial API to advertise Matryoshka-style truncation; Voyage AI added the technique to its general-purpose family with voyage-3-large [2].

Voyage AI's implementation differs from OpenAI's in two practical ways. First, voyage-3-large and the voyage-3.5 models expose a discrete set of supported dimensions (256, 512, 1024, 2048) rather than allowing arbitrary truncation, which lets the API normalize the truncated vector to unit length server-side [2]. Second, Voyage pairs Matryoshka training with quantization-aware training so that the truncated vectors remain accurate not just at float32 but also at int8 and binary precision [2]. The combination enables the headline storage reductions Voyage AI quotes against OpenAI at the same retrieval quality [14]. For application developers, Matryoshka support means a single index can serve multiple query latency profiles: a retrieval pipeline might store 2048-dimensional float32 vectors for offline analytics and re-ranking while serving an online vector search at 256 or 512 int8 dimensions, all from the same API call.

Benchmarks

MTEB leaderboard

Voyage-3-large was the first model in the voyage-3 family to take the top position on the Hugging Face MTEB English leaderboard, which aggregates 56 retrieval, classification, clustering, reranking, summarization, and semantic-similarity tasks [13]. Voyage AI reported the following snapshot of the leaderboard at the launch of voyage-3-large in January 2025 [2].

Model	MTEB average	Provider	Default dim
voyage-3-large	65.1	Voyage AI	1024
NV-Embed v2	64.8	NVIDIA	4096
OpenAI text-embedding-3-large	64.6	OpenAI	3072
stella_en_1.5B_v5	64.4	Independent	1024
Cohere Embed v3 English	64.5	Cohere	1024

MTEB is a general-purpose benchmark, and the gap between the top entries is small. Voyage AI's marketing has therefore emphasized retrieval-specific evaluations on internal datasets covering eight domains, where the company reports larger leads [2].

Voyage AI domain-specific retrieval evaluation

Voyage AI publishes results for its models on a private suite of 100 retrieval datasets grouped into eight domains: technical documentation, code, law, finance, web reviews, multilingual content, long documents, and conversations [2]. The headline numbers, averaged across all eight domains, are summarized below as the percentage uplift over OpenAI text-embedding-3-large (the reference baseline at each launch date).

Model	Avg uplift over OpenAI v3-large	Notes
voyage-3	+7.55%	Reported Sep 2024 on 40-dataset subset [1]
voyage-3-lite	+3.82%	Reported Sep 2024 on 40-dataset subset [1]
voyage-3-large	+9.74%	Reported Jan 2025 on 100-dataset suite; also +20.71% vs Cohere Embed v3 English [2]
voyage-3.5	+8.26%	Reported May 2025; also +2.66% vs voyage-3 and +1.63% vs Cohere Embed v4 [3]
voyage-3.5-lite	+6.34%	Reported May 2025; also +4.28% vs voyage-3-lite [3]

The evaluation is not externally audited, but Voyage AI has released the dataset list and evaluation scripts on GitHub, which has allowed several independent researchers to publish broadly consistent results [2]. Voyage-3-large is best-in-class on every domain except code, where the specialized voyage-code-3 leads [4].

Domain-specialized siblings

The voyage-3 family coexists with several domain-specialized checkpoints that share Voyage AI's encoder lineage but are fine-tuned for narrower use cases. Voyage-code-3 adopts the same Matryoshka and quantization toolkit as the voyage-3-large generation, while the older voyage-law-2 and voyage-finance-2 do not [4].

Model	Released	Domain	Context	Dimensions	Notes
voyage-law-2	Apr 2024	Legal text	16,000	1024	Superseded by voyage-3-large for many uses
voyage-finance-2	May 2024	Financial filings	32,000	1024	Pre-Matryoshka; tuned on SEC and earnings data
voyage-code-3	Dec 2024	Source code	32,000	256, 512, 1024, 2048	Highest scores on code retrieval; Matryoshka-enabled
voyage-multimodal-3	Nov 2024	Text and images	32,000	1024	Single embedding space for text plus image inputs

With voyage-3-large surpassing voyage-law-2 and voyage-finance-2 on most legal and financial retrieval tasks, Voyage AI's practical recommendation as of 2025 is to default to a general-purpose voyage-3 model and fall back to a domain model only when domain-specific recall is critical [2]. Voyage-code-3 remains the recommended model for source-code retrieval and IDE indexing [4].

How does Voyage-3 compare to OpenAI, Cohere, and Google embeddings?

The table below compares the voyage-3 family with the major commercial embedding APIs available in 2025, focusing on default dimension, context window, and the headline price for the standard-quality tier [6]. The price column lists U.S. dollars per million input tokens at the time of each model's most recent pricing update.

Provider	Model	Default dim	Matryoshka dims	Context	Price per 1M tokens
Voyage AI	voyage-3-large	1024	256, 512, 1024, 2048	32,000	$0.18
Voyage AI	voyage-3.5	1024	256, 512, 1024, 2048	32,000	$0.06
OpenAI	text-embedding-3-large	3072	arbitrary truncation	8,192	$0.13
OpenAI	text-embedding-3-small	1536	arbitrary truncation	8,192	$0.02
Cohere	Cohere Embed v3 English	1024	no	512	$0.10
Cohere	Embed v4	1536	256, 512, 1024, 1536	128,000	$0.12
Jina AI	Jina Embeddings v3	1024	32 to 1024	8,192	$0.02
Nomic	Nomic Embed text v1.5	768	64, 128, 256, 512, 768	8,192	open weights
Google	gemini-embedding-001	3072	768, 1536, 3072	2,048	$0.0001 per request

The voyage-3 family stands out for combining a 32,000-token context window with a small default vector size and Matryoshka truncation down to 256 dimensions. OpenAI text-embedding-3-large supports a larger output dimension and arbitrary truncation but uses a shorter 8,192-token context window and produces a much larger default vector [1]. Cohere Embed v4 introduced Matryoshka and a much longer context but at a higher price than voyage-3.5 [3]. Open-weight models such as Nomic Embed v1.5 and Jina Embeddings v3 compete primarily on price and licensing flexibility rather than peak retrieval accuracy.

Anthropic integration and Claude embeddings

When Anthropic added an Embeddings section to its Claude developer documentation in 2024, it stated that Anthropic does not train its own embedding model and that Voyage AI is its recommended embedding provider for Claude-based information retrieval and retrieval-augmented generation [9]. The Anthropic Cookbook on GitHub includes a third-party Voyage AI notebook demonstrating how to chunk documents, embed them with voyage-3 or voyage-3-large, store the vectors in a vector database, and retrieve the top results before passing them to Claude for generation [11].

Did Anthropic acquire Voyage AI?

No. The relationship between Anthropic and Voyage AI is a partnership and a recommendation, not a corporate combination; Anthropic has never acquired Voyage AI. The acquisition that did occur was by MongoDB, which bought Voyage AI for 220 million dollars in a cash-and-stock deal announced on 24 February 2025, roughly 17 months after the startup was founded [7][20]. Voyage AI now operates as Voyage AI by MongoDB, and founder Tengyu Ma became MongoDB's chief AI scientist [8][20]. "For AI applications to reach their full potential, businesses must trust their outputs, so retrieval needs to be deeply integrated with operational data to be accurate and relevant," Ma said at the time of the deal [7]. The companies remained closely associated commercially even after the acquisition, with Anthropic's recommendation language for Voyage embeddings preserved in the Claude documentation [9], and MongoDB stated that Voyage AI's models "will remain available through voyage.ai, AWS Marketplace, and Azure Marketplace" [7].

Anthropic has consistently described its Claude family of models as best paired with strong external retrieval. In its September 2024 research note "Introducing Contextual Retrieval," Anthropic reported that contextual embeddings alone reduced the top-20-chunk retrieval failure rate by 35 percent, that combining them with a contextual BM25 sparse retriever cut the failure rate by 49 percent (from 5.7 percent to 2.9 percent), and that adding a reranking step lifted the reduction to 67 percent [10]. Anthropic tested several embedding providers for that work and wrote, "We found Gemini and Voyage embeddings to be particularly effective" [10].

Architecture and training

Voyage AI does not publish parameter counts or full architectural details for its production embedding models. The company has stated that the voyage-3 family uses an encoder-style transformer with a sequence length of 32,000 tokens and a projection that produces the default 1024-dimensional output [1][5]. The Matryoshka projection head used in voyage-3-large and the voyage-3.5 models adds a structured loss over nested prefixes of the output vector at training time, with quantization-aware training mixed in for int8 and binary precision [2].

The training mixture comprises more than two trillion tokens of multilingual web text, code, mathematics, scientific publications, legal documents, financial filings, and conversational data, augmented with retrieval pairs mined from open datasets and from proprietary sources [1]. Voyage AI applies retrieval-result alignment via human feedback, in which annotators rank candidate retrievals against queries and the model is updated to prefer the higher-ranked candidates [1]. The voyage-3.5 step adds distillation from voyage-3-large to lift the smaller checkpoints without increasing inference cost [3]. All models in the family are trained with contrastive learning over query-document pairs, combining an in-batch contrastive loss, a triplet loss with hard negatives mined from a retriever-in-the-loop process, and the Matryoshka prefix loss.

Availability and pricing

Voyage AI provides the voyage-3 family through its REST API at api.voyageai.com, with Python and JavaScript SDKs [5]. The same models are distributed through several cloud marketplaces: AWS Marketplace as inference endpoints, Microsoft Azure AI Foundry, and Vercel AI Gateway [15][16]. Selected checkpoints, including the voyage-3-m-exp experimental research release, are published on Hugging Face under a non-commercial license, but the production-grade models are API-only. After the MongoDB acquisition, the models were integrated into MongoDB Atlas Vector Search through a managed embedding pipeline that automatically converts new documents into voyage-3 or voyage-3.5 embeddings during ingest, billed at the same per-token rate as the public API, with the free-token allowance now tied to the current voyage-4 generation [6][8].

What replaced Voyage-3? The voyage-4 family

Voyage AI introduced the voyage-4 family on 15 January 2026 as the successor to the voyage-3 generation [18]. The launch comprised four models sharing what Voyage AI called an "industry-first" shared embedding space, meaning all four produce compatible vectors so that a smaller, cheaper model can embed queries against documents embedded by a larger model, a pattern known as asymmetric retrieval [18].

Model	Role	Default dim	Other dims	Price per 1M tokens	Notes
voyage-4-large	Flagship	1024	256, 512, 2048	$0.12	First production embedding model built on a mixture-of-experts architecture
voyage-4	Mid-tier	1024	256, 512, 2048	$0.06	Approaches voyage-3-large quality at mid-size cost
voyage-4-lite	Entry	1024	256, 512, 2048	$0.02	Approaches voyage-3.5 quality with far fewer parameters
voyage-4-nano	Open weight	1024	256, 512, 2048	Apache 2.0	First open-weight Voyage model, on Hugging Face

Voyage AI described voyage-4-large as "the first production-grade embedding model that utilizes a mixture-of-experts architecture" and reported that it surpasses voyage-3-large by 1.87 percent and OpenAI text-embedding-3-large by 14.05 percent on average, and also leads Cohere Embed v4 and Google's Gemini Embedding 001, while the MoE design keeps serving costs about 40 percent lower than a comparable dense model [18]. Notably, voyage-4-large is priced at $0.12 per million tokens, one-third less than voyage-3-large's $0.18, and voyage-4-nano is the company's first open-weight release, distributed under the Apache 2.0 license [6][18].

On 29 June 2026, Voyage AI added voyage-context-4, a contextualized-chunk embedding model that processes an entire document in a single forward pass so that each chunk vector "encodes not only the chunk's own content but also the contextual information from the full document," priced at $0.12 per million tokens [19]. With these releases the voyage-3 family moved into a maintenance role: Voyage AI continues to serve and document it but recommends the voyage-4 models for new projects [6][18].

Reception and impact

Reviewers consistently highlighted three properties of the voyage-3 family: the small default 1024-dimensional vector size, the long 32,000-token context window, and the Matryoshka and quantization toolkit [2]. Independent comparisons by vector database vendors, including Pinecone and Weaviate, reproduced Voyage AI's claimed lead over text-embedding-3-large on retrieval benchmarks while noting that the absolute gap on general English MTEB tasks was small [13]. At the systems-engineering level, the family's 256-dimensional binary embeddings drew particular attention because they enabled fast in-memory nearest-neighbor search on billion-scale corpora with hardware footprints previously associated with classical TF-IDF or BM25 indexes [2].

In the developer community, the family's adoption by Anthropic for Claude-based retrieval-augmented generation, alongside its inclusion in the LangChain and LlamaIndex ecosystems, made voyage-3 a common default for new RAG projects in 2025 [7][10]. The free 200-million-token tier reduced the friction of switching from OpenAI for prototypes, and the MongoDB Atlas integration after the February 2025 acquisition simplified production deployment for users already on the MongoDB stack [8]. By the end of 2025, voyage-3-large and voyage-3.5 were among the most frequently cited commercial embedding models in published RAG research, alongside OpenAI text-embedding-3-large and the open-weight Nomic and Jina families [17].

Limitations and criticisms

The voyage-3 family has attracted several specific criticisms. Voyage AI does not publish parameter counts, training data, or full evaluation scripts, which limits the ability of external researchers to reproduce or audit its claims [2]. The company's domain-evaluation suite of 100 datasets, on which the headline lead over OpenAI and Cohere is reported, is not fully open, and several researchers have argued that the dataset selection favors the kinds of documents Voyage AI fine-tunes on. The MTEB leaderboard, where the absolute lead is much smaller, provides a more neutral comparison [13].

Is Voyage-3 open source?

No, for the production models. The API is proprietary. Unlike Nomic Embed and Jina Embeddings v3, the voyage-3 family is not released as open weights for the production checkpoints, so users cannot self-host or fine-tune the models on their own hardware. The voyage-3-m-exp release on Hugging Face is an experimental research checkpoint with a non-commercial license and is not the same model served by the API. Customers who require on-premises inference have therefore typically continued to use open-weight alternatives. This changed with the next generation: voyage-4-nano, released on 15 January 2026, is the first Voyage model with open weights, distributed under the Apache 2.0 license [18]. With the release of the voyage-4 family the voyage-3 line moved into a maintenance role; Voyage AI continues to serve and document it but recommends voyage-4 for new projects with the highest quality requirements [6][18].

See also

References

Voyage AI. "voyage-3 & voyage-3-lite: A new generation of small yet mighty general-purpose embedding models." 18 September 2024. https://blog.voyageai.com/2024/09/18/voyage-3/
Voyage AI. "voyage-3-large: the new state-of-the-art general-purpose embedding model." 7 January 2025. https://blog.voyageai.com/2025/01/07/voyage-3-large/
Voyage AI. "voyage-3.5 and voyage-3.5-lite: improved quality for a new retrieval frontier." 20 May 2025. https://blog.voyageai.com/2025/05/20/voyage-3-5/
Voyage AI. "voyage-code-3: more accurate code retrieval with lower dimensional, quantized embeddings." 4 December 2024. https://blog.voyageai.com/2024/12/04/voyage-code-3/
Voyage AI Documentation. "Text Embeddings." https://docs.voyageai.com/docs/embeddings
Voyage AI Documentation. "Pricing." https://docs.voyageai.com/docs/pricing
MongoDB, Inc. "MongoDB Announces Acquisition of Voyage AI to Enable Organizations to Build Trustworthy AI Applications." Investor news release, 24 February 2025. https://investors.mongodb.com/news-releases/news-release-details/mongodb-announces-acquisition-voyage-ai-enable-organizations
MongoDB. "Redefining the Database for AI: Why MongoDB Acquired Voyage AI." Company blog, 24 February 2025. https://www.mongodb.com/company/blog/news/redefining-database-ai-why-mongodb-acquired-voyage-ai
Anthropic. "Embeddings." Claude API documentation. https://docs.anthropic.com/en/docs/build-with-claude/embeddings
Anthropic. "Introducing Contextual Retrieval." Anthropic research note, September 2024. https://www.anthropic.com/news/contextual-retrieval
Anthropic Cookbook. "How to create embeddings with Voyage AI." GitHub repository. https://github.com/anthropics/anthropic-cookbook/blob/main/third_party/VoyageAI/how_to_create_embeddings.md
Kusupati, Aditya, et al. "Matryoshka Representation Learning." arXiv:2205.13147, NeurIPS 2022.
Muennighoff, Niklas, et al. "MTEB: Massive Text Embedding Benchmark." arXiv:2210.07316. Leaderboard at https://huggingface.co/spaces/mteb/leaderboard
MongoDB Blog. "Matryoshka Embeddings: Smarter Embeddings With Voyage AI." https://www.mongodb.com/company/blog/technical/matryoshka-embeddings-smarter-embeddings-with-voyage-ai
Vercel AI Gateway. "voyage-3-large model specs and pricing." https://vercel.com/ai-gateway/models/voyage-3-large
AWS Marketplace. "voyage-3 Large Embedding Model." https://aws.amazon.com/marketplace/pp/prodview-ohiz4nvrdvuhq
SiliconANGLE. "MongoDB acquires embedding model provider Voyage AI." 24 February 2025. https://siliconangle.com/2025/02/24/mongodb-acquires-embedding-model-provider-voyage-ai/
Voyage AI. "The Voyage 4 model family: shared embedding space with MoE architecture." 15 January 2026. https://blog.voyageai.com/2026/01/15/voyage-4/
Voyage AI. "voyage-context-4: stop worrying about chunking with our best-performing model." 29 June 2026. https://blog.voyageai.com/2026/06/29/voyage-context-4/
Bloomberg. "MongoDB Buys Voyage AI for $220 Million to Bolster AI Search." 24 February 2025. https://www.bloomberg.com/news/articles/2025-02-24/mongodb-buys-voyage-ai-for-220-million-to-bolster-ai-search
TechCrunch. "Voyage AI is building RAG tools to make AI hallucinate less." 3 October 2024. https://techcrunch.com/2024/10/03/voyage-ai-is-building-rag-tools-to-make-ai-hallucinate-less/

Editing: Voyage-3