TL;DR — Cost Winner by Scale
🏆 Under 5M vectors SUPABASE WINS
Supabase Pro at $25/month includes unlimited queries. Pinecone and Weaviate charge per query on top of storage — Supabase wins at low-to-medium query volume.
⚡ 5M–50M vectors — pay-as-you-go
Pinecone Serverless edges ahead if your query volume is moderate. Qdrant Cloud becomes competitive at higher vector counts due to its straightforward GB pricing.
🚀 50M+ vectors at scale
Self-hosted Qdrant on spot/preemptible VMs dominates — roughly $0.01-0.03/GB versus $0.33+ on managed services. Requires DevOps investment.
2026 Pricing Structure Decoded
Every vector database has a different pricing model, which makes comparison hard. Here's each one decoded to the same units:
Pinecone (Serverless)
- Storage: $0.33 per GB per month
- Reads: $16 per million read units (1 RU ≈ fetching 1 vector with 100 neighbors)
- Writes: $2 per million write units
- Base fee: None (pure pay-as-you-go)
Supabase (pgvector on PostgreSQL)
- Storage: $0.125 per GB per month (database storage)
- Queries: Included in plan — no per-query charge
- Base fee: $25/month (Pro plan — free tier has limits)
Qdrant Cloud
- Storage: ~$9 per GB per month (managed, includes RAM-mapped index)
- Queries: Included
- Base fee: None, but minimum cluster size applies
Weaviate Cloud
- Storage: $0.50 per GB per month
- Queries: $10 per million queries
- Base fee: $25/month (Starter)
Cost at Real Production Scale
All scenarios use 1536-dimension float32 vectors (OpenAI text-embedding-3-small output). For the cost of generating those vectors, see the embedding API cost guide. Index overhead is ~1.5×.
Scenario 1: 1 Million Vectors, 500K queries/month
| Provider | Storage | Query Cost | Base | Monthly Total |
|---|---|---|---|---|
| SupabaseCHEAPEST | $1.13 | $0 | $25 | $26.13 |
| Pinecone | $2.97 | $8.00 | $0 | $10.97 |
| Qdrant | $82.80 | $0 | $0 | ~$83 |
| Weaviate | $4.50 | $5.00 | $25 | $34.50 |
Winner: Pinecone at low query volume. Supabase wins once queries exceed ~2M/month (where Pinecone's $16/M read unit cost overtakes the $25 base fee advantage).
Scenario 2: 10 Million Vectors, 5M queries/month
| Provider | Storage | Query Cost | Base | Monthly Total |
|---|---|---|---|---|
| SupabaseCHEAPEST | $11.25 | $0 | $25 | $36.25 |
| Pinecone | $29.70 | $80.00 | $0 | $109.70 |
| Qdrant | $828 | $0 | $0 | ~$828 |
| Weaviate | $45.00 | $50.00 | $25 | $120 |
Qdrant Cloud's $9/GB managed price looks expensive but it includes RAM-mapped index for ultra-fast nearest-neighbor search. For high-throughput, latency-critical apps it can justify the cost. Self-hosted Qdrant on a $80/month VM can hold the same 10M vectors.
Scenario 3: 100 Million Vectors, 20M queries/month
| Provider | Est. Monthly Cost | Notes |
|---|---|---|
| SupabaseCHEAPEST MANAGED | ~$150 | Needs $175+ plan; pgvector perf degrades at this scale |
| Pinecone | ~$900 | Storage + $320 query cost |
| Weaviate | ~$750 | Storage + $200 query cost |
| Self-hosted Qdrant | ~$150-300 | 2–4 nodes on spot VMs; requires ops team |
Supabase pgvector uses IVFFlat or HNSW indexes inside PostgreSQL. At 50M+ vectors, query latency can degrade without careful index tuning, partitioning, and hardware scaling. Beyond 20M vectors, evaluate Pinecone or a self-hosted purpose-built solution.
🧊 Calculate your exact vector DB cost
Enter your vector count, dimensions, and monthly query volume — get a live side-by-side estimate for all 4 providers.
Open Vector DB Calculator →Beyond Cost: Feature Comparison
| Feature | Pinecone | Supabase | Qdrant | Weaviate |
|---|---|---|---|---|
| Managed / Serverless | ✅ Fully | ✅ Fully | ✅ Cloud | ✅ Cloud |
| Self-hostable | ❌ | ✅ | ✅ Best-in-class | ✅ |
| Hybrid search (vector + keyword) | ✅ | ✅ via FTS | ✅ | ✅ |
| Metadata filtering | ✅ | ✅ SQL WHERE | ✅ Advanced | ✅ |
| Multi-tenancy | ✅ Namespaces | ✅ RLS | ✅ Collections | ✅ |
| Max vectors (managed) | Unlimited | ~50M practical | ~100M+ | Unlimited |
| Query latency (p99) | <100ms | 10-200ms | <50ms | <100ms |
Which Vector Database Should You Choose?
Choose Supabase pgvector if:
- You already use Supabase for your database (zero extra infra)
- Your vector count is under 5M and queries under 5M/month
- You want SQL-native filtering and joins (no impedance mismatch)
- Budget is tight and predictable costs matter more than raw performance
Choose Pinecone if:
- You need a fully managed, zero-ops vector DB as a service
- Your query volume is moderate (under 2M/month) where costs stay reasonable
- You need strong SLAs and enterprise support
Choose Qdrant if:
- You need maximum query performance (lowest p99 latency)
- You're comfortable self-hosting and have DevOps capacity
- You're at 50M+ vectors where self-hosting becomes dramatically cheaper
- You need advanced payload filtering with complex queries
Choose Weaviate if:
- You need built-in multimodal search (text + image + audio)
- You want GraphQL-native API design
- Your team values module ecosystem (Cohere, OpenAI integrations)