TurboQuant
DB
Algorithm
Benchmarks
Code
Advisor
GitHub ↗
Config Advisor
Pick the right config
in seconds.
Scored against real benchmark data across 6 datasets, 1–4 bits, brute and ANN.
① What are you building?
⌬
RAG / Chatbot
Q&A over docs
⌕
Search at scale
1M+ vectors
⌂
Laptop / Edge
Disk & RAM matter
◎
Max accuracy
Recall first
◉
Balanced
General prod use
⏵
Fast ingest
Write throughput
② Embedding dimension
96
200
256
768
1536
3072
Benchmark dataset:
dbpedia-1536
③ Recall@k
k=1
k=2
k=4
k=8
k=16
k=32
how many neighbors retrieved before reranking
Recommended for you
Loading benchmark data…
Tune the weights manually
— override scenario priorities
Recall
50%
Compression
35%
Speed
15%