Menu

Post image 1
Post image 2
Post image 3
1 / 3
0

Postgres With pgvector vs Pinecone: 1 Million Embeddings, One Honest Comparison

DEV Community·Gabriel Anhaia·about 1 month ago
#fIywvwMK
Reading 0:00
15s threshold

Book: Database Playbook: Choosing the Right Store for Every System You Build Also by me: RAG Pocket Guide: Retrieval, Chunking, and Reranking Patterns for Production My project: Hermes IDE | GitHub — an IDE for developers who ship with Claude Code and other AI coding tools Me: xgabriel.com | GitHub You have 1.2 million chunks, 1536 dimensions, a Postgres cluster you already pay for, and a Slack thread three engineers deep arguing about whether to add Pinecone to the bill. Half the team wants the managed service. Half the team is allergic to another vendor. Both halves are quoting blog posts that never agree on the same numbers. This post compares the two on latency, recall, cost per million vectors per month, and operational weight. Where the data comes from a named public benchmark, you get the link. Where the data is your-team-shaped, it is clearly labeled illustrative — because most of the cost-per-team numbers floating around online are fiction in a costume.…

Continue reading — create a free account

Join HashtagPLUS to read full articles, follow hashtags, vote, and join the conversation.

Read More