infino bench

full-text 6.20 ms BM25 keyword search, top 10 $0.14 / 1M queries
vector 13.81 ms 10M × 1024-dim, top 10 $0.31 / 1M queries
hybrid 16.57 ms keyword + vector, fused, top 10 $0.38 / 1M queries

infino bench --scale 1M..1B

  • 1M 55 µs 7.9 ms
  • 10M 123 µs 49 ms
  • 100Mazure 9.6 ms 745 ms
  • 1Bazure 97 ms 6.31 s
full table: ingest rate, stored size, cold start
docsstoreingeststored (ratio)warm rangecold open
1MAWS S338.7K docs/sn/a55 µs to 7.9 ms0.42-0.56 s
10MAWS S376.9K docs/s6.17 GiB (33%)123 µs to 49 ms0.45-0.51 s
100MAzure Blob177.1K docs/s59.33 GiB (32%)9.6 ms to 745 ms0.55-1.10 s
1BAzure Blob110.6K docs/s593.31 GiB (32%)97 ms to 6.31 s4.8-9.5 s

infino cost

dataingest rateraw → storedratiostorage
full-text76.9K docs/s18.72 GiB → 6.17 GiB33%$0.15/mo
vector (1024d)20.2K docs/s38.15 GiB → 21.10 GiB55%$0.52/mo
sql50.4K rows/s18.72 GiB → 6.96 GiB37%$0.17/mo

→ model your own workload in the cost calculator

infino bench fts

  • point lookup 123 µs
  • typical query 3.30 ms
  • broad, worst case 48.98 ms
full table: every query shape, warm, cold, and cost
querywarmwarm + fetchcold opencold search$ / 1M
single term · df=1123.44 µs3.92 ms450.23 ms11.49 ms$0.0039
single term · rare1.11 ms9.88 ms476.38 ms183.71 ms$0.03
single term · common3.30 ms11.04 ms487.34 ms303.69 ms$0.10
2-term OR1.35 ms9.72 ms498.61 ms402.66 ms$0.04
2-term AND1.81 ms10.01 ms484.03 ms391.72 ms$0.06
3-term OR (wide)2.71 ms10.63 ms505.87 ms440.47 ms$0.09
3-term AND (wide)3.63 ms11.48 ms493.72 ms411.78 ms$0.11
5-term OR15.55 ms23.57 ms492.02 ms350.97 ms$0.49
5-term AND7.62 ms15.97 ms486.09 ms348.38 ms$0.24
10-term OR48.98 ms57.05 ms475.07 ms337.18 ms$1.54
10-term AND9.35 ms17.54 ms489.39 ms439.95 ms$0.29

infino bench vector

warm search
15.56 mssteady state, median
cold start
1.09 s + 1.42 sopen the index once, then the first search
recall
0.96+fraction of true nearest neighbors returned, vs exact search
stored size
21.10 GiB55% of the 38.15 GiB raw vectors

infino bench hybrid

functionwarmresultsrelative
bm25_search() 6.20 ms 10
vector_search() 13.81 ms 10
hybrid_search() 16.57 ms 10
token_match() · one row 381 µs 1
exact_match() 2.81 ms 1

infino bench --env

machine
one Intel Xeon Platinum 8488C · 8 cores / 16 threads · 31 GiB RAM · Linux
cluster
none, a single node
data layout
256 Parquet files on object storage, read through a disk cache
corpus
synthetic, realistic word-frequency skew (Zipfian), 200 words per doc
measurement
median (p50) of repeated runs, one query at a time
harness
in-tree, one binary, open source: build once, then measure warm, then cold

the runs on this page

scalevectors
10M docs1024-dim vectors
1M docs384-dim vectors
100M and 1B (full-text)n/a