AITOT

Kalkulator

Estimator Biaya Vector DB

Perkirakan biaya bulanan berdasarkan jumlah vektor, dimensi, dan query/hari.

Harga diperbarui:

AITOT Vector DB Cost Estimator membandingkan biaya bulanan di 9 provider — Pinecone Serverless/Pod, Qdrant Cloud, Weaviate Cloud, Milvus Cloud, Supabase pgvector, Turso, Chroma Cloud, dan self-hosted.

Masukkan vector count, dimensi embedding, dan queries per hari. Kalkulator surface storage, query, dan index overhead terpisah.

Toggle binary quantization (didukung Pinecone, Qdrant, Milvus) untuk lihat kompresi storage 32×. Vector 1536-dim kompres dari 6KB ke 192 byte.

Termurah

LanceDB Cloud · Standard

Ukuran terindeks: 6.87 GB (Ukuran mentah: 5.72 GB)

$6.99

/bulan

ProviderPenyimpananBacaTulisTotal / bulan
LanceDB CloudStandard$0.69$6.00$0.30$6.99
TurbopufferObject-store backed (Launch)

S3-backed; cold reads slower but radically cheaper at scale. Launch plan min $16/mo (was $64/mo pre-2026-06).

$0.25$4.50$0.22$16.00minimum
Supabase pgvectorPro (8GB included)

Bundled with Postgres compute; query cost folded into DB hours

$1.07——$25.00minimum
PineconeServerless (s1, Standard)

us-east-1; reads billed in Read Units (1 RU ≈ 1 query × 1KB result). Standard plan $50/mo min usage.

$2.64$24.00$0.60$27.24
Zilliz / MilvusServerless

5M free vectors; pay-as-you-go above. vCU-based billing ($4/M vCU); read/write $ are derived floors, not flat per-op rates.

$0.30$36.00$0.45$36.75
Weaviate CloudFlex (formerly Serverless)

Flex plan (renamed from Serverless, 2026-07). New dimension-based billing model — storage/read/write split below is a legacy approximation, verify-pending.

$0.32$7.50$0.38$45.00minimum
MongoDB AtlasM10 + Vector Search

Vector search included with Atlas tier at no extra software cost

$2.00——$57.00minimum
Qdrant CloudHybrid Cloud (1×1GB)

Per-node billing ~$76/node; small nodes hold ~30 GB indexed

$18.60——$76.00minimum
PineconePod (p1.x1)

Legacy: 1 pod ≈ 5M 768-dim vectors; scale by adding pods (~$81/pod-mo Standard). New users default to Serverless.

$129.78——$129.78
PineconeServerless (s1, Enterprise)

Enterprise tier: higher SLA, higher RU/WU rates than Standard, $500/mo min usage.

$2.64$36.00$0.90$500.00minimum

Overhead index dan harga bervariasi per region. Kuantisasi (int8/binary) menukar recall dengan biaya — selalu benchmark sebelum deploy.

Yang dilakukan kalkulator ini

9 vector DB dibandingkan

Pinecone Serverless + Pod, Qdrant, Weaviate, Milvus, Supabase pgvector, Turso, Chroma, self-host.

Binary quantization

Toggle untuk 32× kompresi storage.

Scaling dimensi

Lihat 1536 vs 3072 dimensi pengaruhi storage + query cost.

Storage + query split

Biaya dipisah jadi index storage, query operations, metadata.

Break-even self-host

Bandingkan managed dengan VPS $80/bulan Qdrant atau Milvus.

Hybrid + filter pricing

Termasuk biaya filter operation.

Perbandingan cepat

Biaya bulanan di 10M × 1536-dim vektor, 100k queries/hari

ProviderStorageQueriesTotal / bulan
Supabase pgvector (Pro)incl.incl.$25
Qdrant Cloud (basic)$30$10$40
Weaviate Serverless$35$10$45
Milvus Cloud (Starter)$50$15$65
Pinecone Serverless$50$40$90
Chroma Cloud$70$20$90
Pinecone Pod-based (s1)$100$40$140
Self-host di VPS $80$80incl.$80

Biaya turun 60-80% dengan binary quantization.

Cara menggunakan kalkulator

Bandingkan biaya bulanan di 9 vector database untuk corpus dan query volume Anda.

  1. 1

    Masukkan vector count

    Dokumen × chunks/dok. Tipikal: 1 dok = 5-20 chunks @ 500 token.

  2. 2

    Pilih dimensi

    1536 default OpenAI. 3072 gandakan storage; 768 setengah.

  3. 3

    Set query volume

    Queries harian. Caching kurangi — mayoritas cache 30-50%.

  4. 4

    Toggle binary quant

    Di Pinecone, Qdrant, Milvus, binary quant potong storage 32×.

Kenapa pakai kalkulator ini

  • ✓9 provider di-track bulanan
  • ✓Modeling binary quantization
  • ✓Perbandingan self-host
  • ✓Biaya dimensi + filter terpisah
  • ✓Free tier limits ditampilkan
  • ✓Tanpa login

Pertanyaan yang sering diajukan

Vector DB mana termurah di 10M vektor 2026?+
Untuk 10M × 1536-dim vektor di volume query sedang: pgvector di Supabase Pro $25/bulan termurah. Qdrant Cloud $40/bulan dan Weaviate Serverless $45/bulan mengikuti. Pinecone Serverless $90/bulan dan Pinecone Pod-based $140/bulan premium.
Pinecone Serverless vs Pod-based pricing?+
Serverless: $0.40/M writes + $4 per juta reads + $0.33/GB-bulan storage. Pod-based: $0.096/jam untuk s1 starter (~$70/bulan base). Serverless menang di query volume rendah dan traffic bursty. Pod-based menang di >500 reads/detik steady-state.
Apakah dimensi vektor sangat memengaruhi biaya?+
Ya. Dari 1536 (OpenAI text-embedding-3-large) ke 3072 (Cohere Embed v4) menggandakan storage di mayoritas provider. Dengan binary quantization (Pinecone, Qdrant), vektor 1536-dim terkompresi jadi 192 byte — 32× lebih kecil. Cek toggle "use binary quantization".
Pinecone vs Weaviate vs Qdrant — pilih mana 2026?+
Pinecone: ops termudah, harga premium. Weaviate: terbaik untuk hybrid search + graph. Qdrant: termurah di skala, filtering terbaik. Supabase pgvector: terbaik kalau sudah ada Postgres. Kalkulator menampilkan biaya berdampingan di skala Anda.
Cara potong tagihan vector DB jadi setengah?+
Tiga tuas terbesar: (1) kurangi dimensi via Matryoshka — simpan di 512 alih-alih 1536 dengan loss recall minimal, (2) aktifkan binary quantization, (3) pindah dari Pinecone Pod-based ke Qdrant Cloud atau self-host VPS $80/bulan untuk hemat 80%.
Self-host Qdrant atau Milvus benar lebih murah?+
Di 10M+ vektor, ya — VPS $80/bulan jalankan Qdrant fine untuk 50M vektor. Di bawah 5M vektor, managed (Pinecone Free, Qdrant Cloud free) sering menang total cost ketika dihitung waktu ops. Kalkulator menyertakan estimasi self-host.