Vector databases for enterprise buyers
Asked as “vector database”, and as “vector search database for AI applications”, on behalf of an enterprise B2B company. 54 first choices recorded across the direct, paraphrase, budget and scale prompts, twelve models each.
Standing · first-choice share
59%
Clear leader
59Pinecone09Milvus09Qdrant22others
59% of first choices, clear leader.
By buyer segment
The same question asked on behalf of a different buyer. Each standing is computed within its segment; they sit side by side and are never added together.
The standing
Share is the count of first choices across the direct, paraphrase, budget and scale prompts, over all twelve models, for an enterprise B2B company. Ordered by share.
ProductFirst-choice shareNegative rateLabelsQuadrant
Show the three products at 0%, ordered by negative rate
Bars are the share of first choices, 0 to 100Every product with at least 10 labels here. Every product name links to its product page.
All fifteen head-to-head pages: the top six products, each against each
One product takes 59% of first choices here, so the chart would put nine markers in one corner and one at the far edge. The two measurements it plots are columns in the standing above: share, and the negative label rate. One product carries a negative rate above 25% in this category.
What they warned about
Seven of twelve models held their first choice under the paraphrase. Claude Haiku 4.5, Perplexity Sonar, Mistral Small, Llama 4 Maverick and MiniMax M2.5 changed. A high negative share on a product with few labels is a warning. A low share on a product with many labels is salience, not sentiment.
Chroma52%
12 of 23 labels negative · 11 of 12 models · 7 hard negative
“"Avoid Chroma OSS/local-only deployments when: You need mature enterprise governance, strict uptime, multi-region HA..."; "it is not a production vector database for multi-user workloads"” Muse Glimmer 30B, negative prompt
Pinecone25%
19 of 77 labels negative · 11 of 12 models · 5 hard negative
“**High-Risk Vendors to Avoid** ... **Why Avoid:** Despite raising $138M at a $750M valuation, the company reportedly lost a major customer and is shopping for an exit” GLM 4.7 FlashX, negative prompt
Milvus23%
13 of 57 labels negative · 10 of 12 models · 2 hard negative
“Avoid self-hosting **Milvus** unless you have Kubernetes experts; prefer a managed alternative.” Qwen 3.7 Flash, negative prompt
Weaviate13%
9 of 71 labels negative · 8 of 12 models
“Weaviate and Qdrant both document multitenancy features, but the burden is still on the enterprise to design them correctly and validate controls.” GPT-5.4 mini, negative prompt
What they cite
Citations exist only for the models that return a source list: fourteen of the twelve in this edition, and all six flagship models on the expanded tier.
Sites the answers cite
78 of 84 answers in this category came back with a source list, from 14 of 12 models: citations where the model returns them, or the search results it consulted. 1083 links across 258 sites, every framing counted. Ranked by the number of answers carrying the site or page.
vendor site · Atlan42 answers · 48 citations · 12 models
38 answers · 49 citations · 12 models
35 answers · 46 citations · 11 models
25 answers · 26 citations · 9 models
vendor site · G224 answers · 28 citations · 11 models
22 answers · 22 citations · 10 models
21 answers · 21 citations · 10 models
vendor site · Firecrawl17 answers · 18 citations · 8 models
17 answers · 17 citations · 9 models
vendor site · Pinecone15 answers · 32 citations · 7 models
15 answers · 21 citations · 10 models
15 answers · 17 citations · 9 models
Pages the answers cite
The ten pages named in the most answers, by full address. A page here is one the models returned with a recommendation, not one the index endorses.
42 answers · 12 models
38 answers · 12 models
25 answers · 9 models
24 answers · 11 models
22 answers · 10 models
22 answers · 9 models
21 answers · 10 models
17 answers · 8 models
17 answers · 9 models
15 answers · 8 models
Search against answers
Each company's standing in the answers beside its site's footprint in Google search, one row a site: the products the models named on it with their shares, and the share they add up to; monthly searches on Google, and DataForSEO's estimate of AI search demand (modeled from search signals, directional, not a count of queries to any assistant), for the most-searched of the company's and its products' names (the name is in each row's hover text); estimated monthly organic visits to the site; and its best position in Google's top ten for “best vector database”, “vector database”, “vector databases”. US estimates from DataForSEO and Google's Ads Transparency Center. A small company's site, or a mid-sized company's site for its flagship, is marked company; a product on a large parent's site (Google, Microsoft) has no site figures. A column with no figures for this category is left out, and an empty cell means none were seen, not none exist. Two measurements side by side: neither is read as the cause of the other.
| Company and products | Share | Own site cited | Name searches, Google | AI search demand, est. | Organic visits | Google ads, last 30 days | Best Google position |
|---|---|---|---|---|---|---|---|
| Pinecone pinecone.io | 59% | 201 | 60,500 | 3,929 | 90,633 | 14 | 4 |
| Qdrant qdrant.tech | 9% | 198 | 6,600 | 180 | 8,064 | 24 | |
| Milvus milvus.io Milvus 9%, Milvus / Zilliz Cloud 0% | 9% | 127 | 3,600 | 169 | 52,360 | ||
| Weaviate weaviate.io | 7% | 187 | 2,900 | 54 | 5,054 | ||
| pgvector | 7% | 5,400 | 119 | ||||
| Zilliz Cloud zilliz.com | 2% | 28 | 140 | 2 | 11,801 | 24 | 7 |
| Elasticsearch | 2% | 12,100 | 2,798 | ||||
| Elastic elastic.co Elastic 2%, Elasticsearch / OpenSearch 0% | 2% | 16 | 74,000 | 34,672 | 208,306 | 105 | |
| Self-Hosted Qdrant | 2% | ||||||
| Azure AI Search | 0% | 301,000 | 47,904 |
Google's top ten for the three searches hold 16 sites; 3 of them are among the sites the models cited here (encore.dev, g2.com, pinecone.io). The first result for “best vector database” is zenml.io.
How they answered
Six framings of the same buying question, each sent to every model in a fresh session with search on. One row per model, so a row shows whether it held its answer under rewording, what it named when cost was the constraint, and what it argued against. Computed from the raw judge labels.ShowHide
How they answered
Six framings of the same buying question, each sent to every model in a fresh session with search on. One row per model, so a row shows whether it held its answer under rewording, what it named when cost was the constraint, and what it argued against. Computed from the raw judge labels.
ShowHideModelDirect“What is the best vector database for an enterprise B2B company?”Paraphrase“Which vector search database for AI applications would you recommend to a large B2B company with thousands of employees?”Comparative“What are the top enterprise-grade vector databases and how do they differ?”Budget-constrained“What is the best vector database for a large company that needs predictable total cost across thousands of users?”Scale-constrained“We are a 5,000 person company with SSO, SOC 2 and procurement review requirements evaluating a vector database. What should we look at?”Negative“Which vector databases should a large enterprise avoid or be cautious about?”
Claude Haiku 4.5no first choiceElasticsearch, PineconeChangedPineconePineconeno first choiceagainst: Chroma, FAISS, Marqo, Pinecone, Qdrant, Redis Vector Search, Weaviate
Three alternatives
Milvus, Weaviate, Zilliz CloudFive alternatives
Azure AI Search, Milvus, Qdrant, Weaviate, pgvectorOne alternative
QdrantGPT-5.4 miniPineconePineconeHeldPineconeagainst: Elasticsearch, Redis Vector SearchWeaviateagainst: Milvus, Qdrantno first choiceagainst: Milvus, Qdrant, Weaviate, Zvec
Five alternatives
Azure AI Search, MongoDB Atlas Vector Search, OpenSearch, Qdrant, WeaviateTwo alternatives
Milvus, WeaviateFour alternatives
Milvus, Qdrant, Weaviate, pgvectorOne alternative
PineconeGemini 3.5 FlashpgvectorpgvectorHeldPinecone, pgvectoragainst: SingleStoreWeaviateagainst: Pineconepgvectoragainst: Pineconeagainst: Chroma, FAISS, Milvus, Pinecone, SQLite-vec
Five alternatives
Elasticsearch, Milvus, OpenSearch, Pinecone, WeaviateThree alternatives
Elasticsearch / OpenSearch, Milvus, PineconeFour alternatives
Elasticsearch / OpenSearch, Milvus, Qdrant, WeaviateThree alternatives
Milvus, Qdrant, pgvectorSeven alternatives
Databricks, Elasticsearch / OpenSearch, Milvus, Qdrant, Snowflake, Weaviate, Zilliz CloudPerplexity SonarPineconeElastic, PineconeChangedPineconeMilvus, Qdrantagainst: Pineconeno first choiceagainst: pgvector
Four alternatives
Elasticsearch, Milvus, Qdrant, WeaviateFour alternatives
Milvus, Qdrant, Weaviate, Zilliz CloudSeven alternatives
Azure AI Search, Elasticsearch, Milvus / Zilliz Cloud, Qdrant, Redis, Weaviate, pgvectorTwo alternatives
Weaviate, pgvectorGrok 4.20PineconePineconeHeldPineconeagainst: Elasticsearch, Redis, Vespa, pgvectorQdrantagainst: PineconePineconeagainst: Milvus, pgvectoragainst: Chroma, FAISS, LanceDB, Milvus / Zilliz Cloud, Redis, Weaviate, pgvector
Four alternatives
Milvus / Zilliz Cloud, Qdrant, Weaviate, pgvectorFour alternatives
Milvus, Qdrant, Weaviate, pgvectorFour alternatives
Milvus, Qdrant, Weaviate, Zilliz CloudThree alternatives
Milvus, Weaviate, pgvectorTwo alternatives
Qdrant, Zilliz CloudMistral SmallMilvusPineconeChangedPineconeMilvus, Qdrant, Weaviateagainst: Pineconeno first choiceagainst: Chroma, Milvus, Pinecone, Qdrant
Three alternatives
Elasticsearch, Pinecone, WeaviateThree alternatives
Milvus, Qdrant, WeaviateFive alternatives
Milvus / Zilliz Cloud, Qdrant, Vespa, Weaviate, pgvectorDeepSeek V4 FlashPineconePineconeHeldagainst: pgvectorPineconeagainst: Chroma, pgvectorQdrantagainst: Pinecone, pgvectorPineconeagainst: ChromaDB, Milvus, Pinecone, Vespa, Weaviate, Zilliz Cloud
Four alternatives
Milvus / Zilliz Cloud, Qdrant, Weaviate, pgvectorThree alternatives
Milvus / Zilliz Cloud, Qdrant, WeaviateThree alternatives
Milvus, Qdrant, WeaviateOne alternative
MilvusThree alternatives
Milvus, Qdrant, WeaviateLlama 4 Maverickno first choicePineconeChangedno first choiceno first choicePineconeagainst: Chroma, LanceDB, Milvus, Qdrant, Weaviate
Three alternatives
Milvus, Qdrant, WeaviateQwen 3.7 FlashPineconePineconeHeldPineconeagainst: Chroma, pgvector + pgvectorscaleQdrantagainst: PineconePineconeagainst: Milvus, Qdrantagainst: Chroma, Milvus, Pinecone
Two alternatives
Zilliz Cloud, pgvectorTwo alternatives
Elastic, WeaviateThree alternatives
Milvus / Zilliz Cloud, Qdrant, WeaviateOne alternative
Zilliz CloudThree alternatives
Amazon OpenSearch Service, Azure AI Search, ElasticsearchKimi K2PineconePineconeHeldPineconeagainst: pgvectorMilvusagainst: PineconePineconeagainst: Qdrantagainst: Chroma, Pinecone, pgvector
Two alternatives
Milvus / Zilliz Cloud, WeaviateThree alternatives
Azure AI Search, Milvus / Zilliz Cloud, WeaviateThree alternatives
Milvus, Qdrant, WeaviateOne alternative
RedisFour alternatives
Elasticsearch, Milvus, Weaviate, Zilliz CloudGLM 4.7 FlashXPineconePineconeHeldno first choiceZilliz Cloudagainst: Pineconeno first choiceagainst: Pinecone, Weaviate
Four alternatives
Milvus / Zilliz Cloud, Qdrant, Weaviate, pgvectorThree alternatives
Milvus / Zilliz Cloud, Weaviate, pgvectorOne alternative
QdrantMiniMax M2.5Pinecone, pgvectorMilvus, PineconeChangedno first choiceSelf-Hosted Qdrantagainst: Pinecone, Upstash VectorPineconeagainst: Milvus, Weaviateagainst: Chroma, Milvus, Pinecone, Weaviate
Two alternatives
Qdrant, WeaviateThree alternatives
Qdrant, Weaviate, pgvectorFive alternatives
Milvus, Pinecone, Qdrant, Weaviate, pgvectorTwo alternatives
Milvus, WeaviateThree alternatives
Azure AI Search, Elasticsearch, RedisBold is the first choiceAlternatives are counted; the count opens them.What the answer argued against
The record
One row per call: the version string exactly as returned, whether the model searched, sources cited, and latency. Full answer text is in the free responses file. Download the record
Seventy-two rows: every prompt, every model, every answer.
| Prompt | Model | Version string | Time (UTC) | Searched | Sources | Latency |
|---|---|---|---|---|---|---|
| Direct recommendation | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 10:52 | no | 0 | 5 s |
| Direct recommendation | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 09:11 | yes | 6 | 6 s |
| Direct recommendation | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 10:28 | yes | 10 | 17 s |
| Direct recommendation | Perplexity Sonar | sonar | 2026-10-01 08:57 | yes | 17 | 3 s |
| Direct recommendation | Grok 4.20 | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 12:22 | yes | 17 | 7 s |
| Direct recommendation | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 09:00 | yes | 13 | 8 s |
| Direct recommendation | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 10:40 | yes | 24 | 31 s |
| Direct recommendation | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 12:50 | yes | 5 | 2 s |
| Direct recommendation | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 07:51 | no | 0 | 27 s |
| Direct recommendation | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 12:32 | yes | 14 | 39 s |
| Direct recommendation | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 13:03 | yes | 23 | 48 s |
| Direct recommendation | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 08:11 | yes | 7 | 18 s |
| Paraphrase | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 10:24 | yes | 16 | 10 s |
| Paraphrase | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 08:47 | yes | 3 | 5 s |
| Paraphrase | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 10:19 | yes | 9 | 25 s |
| Paraphrase | Perplexity Sonar | sonar | 2026-10-01 12:02 | yes | 19 | 4 s |
| Paraphrase | Grok 4.20 | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 09:49 | yes | 23 | 6 s |
| Paraphrase | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 08:25 | yes | 15 | 8 s |
| Paraphrase | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 13:06 | yes | 25 | 26 s |
| Paraphrase | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 07:53 | yes | 5 | 1 s |
| Paraphrase | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 13:18 | no | 0 | 34 s |
| Paraphrase | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 11:46 | yes | 14 | 19 s |
| Paraphrase | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 08:54 | yes | 18 | 25 s |
| Paraphrase | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 11:28 | yes | 5 | 37 s |
| Comparative | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 10:02 | yes | 9 | 10 s |
| Comparative | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 11:07 | yes | 6 | 8 s |
| Comparative | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 13:09 | yes | 8 | 26 s |
| Comparative | Perplexity Sonar | sonar | 2026-10-01 08:49 | yes | 18 | 4 s |
| Comparative | Grok 4.20 | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 09:59 | yes | 23 | 8 s |
| Comparative | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 08:24 | yes | 10 | 9 s |
| Comparative | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 10:23 | yes | 24 | 39 s |
| Comparative | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 12:41 | yes | 5 | 2 s |
| Comparative | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 08:12 | yes | 15 | 41 s |
| Comparative | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 12:41 | yes | 14 | 29 s |
| Comparative | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 09:19 | yes | 25 | 106 s |
| Comparative | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 12:21 | yes | 15 | 19 s |
| Budget constrained | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 11:12 | yes | 18 | 7 s |
| Budget constrained | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 09:33 | yes | 4 | 5 s |
| Budget constrained | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 13:34 | yes | 22 | 32 s |
| Budget constrained | Perplexity Sonar | sonar | 2026-10-01 13:53 | yes | 19 | 4 s |
| Budget constrained | Grok 4.20 | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 10:20 | yes | 22 | 10 s |
| Budget constrained | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 08:38 | yes | 5 | 3 s |
| Budget constrained | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 09:03 | yes | 22 | 35 s |
| Budget constrained | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 13:24 | yes | 5 | 3 s |
| Budget constrained | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 07:38 | yes | 5 | 34 s |
| Budget constrained | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 08:51 | yes | 23 | 28 s |
| Budget constrained | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 10:36 | yes | 22 | 144 s |
| Budget constrained | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 09:14 | yes | 18 | 52 s |
| Scale constrained | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 13:50 | no | 0 | 6 s |
| Scale constrained | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 13:30 | yes | 3 | 9 s |
| Scale constrained | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 10:04 | no | 0 | 21 s |
| Scale constrained | Perplexity Sonar | sonar | 2026-10-01 10:10 | yes | 19 | 9 s |
| Scale constrained | Grok 4.20 | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 08:10 | yes | 18 | 8 s |
| Scale constrained | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 09:47 | no | 0 | 9 s |
| Scale constrained | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 12:24 | yes | 23 | 52 s |
| Scale constrained | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 13:15 | yes | 5 | 2 s |
| Scale constrained | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 13:49 | yes | 5 | 34 s |
| Scale constrained | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 10:37 | yes | 8 | 41 s |
| Scale constrained | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 08:54 | yes | 18 | 33 s |
| Scale constrained | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 13:21 | yes | 9 | 40 s |
| Negative framing | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 08:55 | yes | 16 | 10 s |
| Negative framing | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 08:58 | yes | 4 | 7 s |
| Negative framing | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 07:53 | yes | 14 | 31 s |
| Negative framing | Perplexity Sonar | sonar | 2026-10-01 09:05 | yes | 21 | 5 s |
| Negative framing | Grok 4.20 | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 10:20 | yes | 21 | 16 s |
| Negative framing | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 09:34 | yes | 5 | 4 s |
| Negative framing | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 13:45 | yes | 24 | 35 s |
| Negative framing | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 11:09 | yes | 5 | 1 s |
| Negative framing | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 11:43 | yes | 24 | 65 s |
| Negative framing | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 12:13 | yes | 23 | 23 s |
| Negative framing | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 09:23 | yes | 20 | 39 s |
| Negative framing | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 08:34 | yes | 23 | 70 s |
Normalization in this category
Every judgment call made between the raw labels and the numbers above, listed so it is visible and reversible.
ShowHide
Category-scoped readings
MongoDB read as MongoDB Atlas Vector Search
MongoDB Atlas read as MongoDB Atlas Vector Search
PostgreSQL read as pgvector
PostgreSQL (`pgvector`) read as pgvector
Unresolved, counted raw
Actian VectorAI
Aurora pgvector
Azure Cognitive
Elasticsearch with vector plugins
MyScaleDB
Self-Hosted Qdrant
Vectara
Zvec
Discontinued, still offered
No shut-down product was recommended here.