Cohere-10M — 2026-06-11

r8i.16xlarge · peak recall 0.9981 @ 30 QPS

QPS at recall

Recall	Meerkat	pgvector HNSW	pgvector HNSW (bit)	Δ vs best competitor
0.90	535	493	210	+9%
0.95	291	283	90	+3%
0.99	84	65	— (max r=0.981)	+29%

Latency at recall

	Meerkat	pgvector HNSW	pgvector HNSW (bit)	Δ vs best competitor
@ recall 0.90
p50	1.84 ms	2.01 ms	4.80 ms	-8%
p95	2.32 ms	2.89 ms	5.84 ms	-20%
p99	2.53 ms	3.49 ms	6.31 ms	-27%
@ recall 0.95
p50	3.43 ms	3.59 ms	11.21 ms	-4%
p95	4.09 ms	4.68 ms	13.44 ms	-13%
p99	4.44 ms	5.30 ms	14.16 ms	-16%
@ recall 0.99
p50	11.97 ms	15.64 ms	—	-23%
p95	13.54 ms	19.74 ms	—	-31%
p99	14.10 ms	20.67 ms	—	-32%

All values are taken at exactly the stated recall: each configuration's own sweep is log-interpolated between its two surrounding measured points (measured values are used verbatim on an exact hit) and the best configuration at that recall is shown — values are never interpolated across different configurations. “—” means no single sweep spans that recall; nothing is extrapolated. Recall is distance-based (ann-benchmarks, tie-aware). Insert/update/delete workloads are not measured.

Interpolated on config — Meerkat @0.90: mkt-n24000-cctrue-fs16-soar-cfs · pgvector HNSW @0.90: hnsw-m24 · pgvector HNSW (bit) @0.90: hnsw-m24-bq-rerank · Meerkat @0.95: mkt-n24000-cctrue-fs16-soar-cfs · pgvector HNSW @0.95: hnsw-m24 · pgvector HNSW (bit) @0.95: hnsw-m24-bq-rerank · Meerkat @0.99: mkt-n24000-cctrue-fs16-soar-cfs · pgvector HNSW @0.99: hnsw-m24

Build comparison

	Meerkat	pgvector HNSW	pgvector HNSW (bit)	Δ vs best competitor
Build time	312 s	2742 s	—	-89%
Index size	2.3 GB	38.1 GB	—	-94%

Build stats are for the configuration backing each column's best operating point: Meerkat = mkt-n24000-cctrue-fs16-soar-cfs · pgvector HNSW = hnsw-m24.

Plots

Pareto svg · Distance-recall svg · Per-config svg · Build times svg · Index sizes svg

Host

Instance type	r8i.16xlarge
Hostname	ip-172-31-71-30
CPU	Intel(R) Xeon(R) 6975P-C (64 vCPU)
Memory	496 GB
OS	Ubuntu 26.04 LTS
Kernel	7.0.0-1006-aws

Versions

rekall	0.1.0 (f5764a8 / main)
PostgreSQL	PostgreSQL 18.4 (Ubuntu 18.4-1.pgdg26.04+1) on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 15.2.0-16ubuntu1) 15.2.0, 64-bit
meerkat	0.1.0 (65df24d)
pg_prewarm	1.2
vector	0.8.2

Search

S3 archive

s3://rekall-bench/runs/cohere-10m/2026-06-11/r8i.16xlarge-224329/

Index builds

Index	Type	Build time	Size	Workers
hnsw-m24	pgvector-hnsw	2742.3s	38.1 GB	32
hnsw-m24-bq	pgvector-hnsw	892.7s	4.6 GB	32
mkt-n12000-ccfalse-fs16-soar	meerkat	511.6s	2.3 GB	8
mkt-n12000-cctrue-fs16-soar	meerkat	516.4s	2.2 GB	8
mkt-n16000-ccfalse-fs16-soar	meerkat	647.0s	2.3 GB	8
mkt-n16000-cctrue-fs16-soar	meerkat	646.8s	2.3 GB	8
mkt-n20000-ccfalse-fs16-soar	meerkat	828.0s	2.4 GB	8
mkt-n24000-ccfalse-fs16-soar	meerkat	923.1s	2.4 GB	8
mkt-n24000-cctrue-fs16-soar-cfs	meerkat	312.0s	2.3 GB	32
mkt-n8000-ccfalse-fs16-soar	meerkat	371.9s	2.2 GB	8
mkt-n8000-cctrue-fs16-soar	meerkat	374.4s	2.2 GB	8

Index parameters

hnsw-m24: {"ef_construction":200,"encoding":"vector","m":24}
hnsw-m24-bq: {"ef_construction":200,"encoding":"bit","m":24}
mkt-n12000-ccfalse-fs16-soar: {"boundary_epsilon":0.05,"centroid_compression":false,"centroid_fastscan":null,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":12000,"rerank_vectors":null,"soar_lambda":1.0}
mkt-n12000-cctrue-fs16-soar: {"boundary_epsilon":0.05,"centroid_compression":true,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":12000,"rerank_vectors":null,"soar_lambda":1.0}
mkt-n16000-ccfalse-fs16-soar: {"boundary_epsilon":0.05,"centroid_compression":false,"centroid_fastscan":null,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":16000,"rerank_vectors":null,"soar_lambda":1.0}
mkt-n16000-cctrue-fs16-soar: {"boundary_epsilon":0.05,"centroid_compression":true,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":16000,"rerank_vectors":null,"soar_lambda":1.0}
mkt-n20000-ccfalse-fs16-soar: {"boundary_epsilon":0.05,"centroid_compression":false,"centroid_fastscan":null,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":20000,"rerank_vectors":null,"soar_lambda":1.0}
mkt-n24000-ccfalse-fs16-soar: {"boundary_epsilon":0.05,"centroid_compression":false,"centroid_fastscan":null,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":24000,"rerank_vectors":null,"soar_lambda":1.0}
mkt-n24000-cctrue-fs16-soar-cfs: {"boundary_epsilon":0.05,"centroid_compression":true,"centroid_fastscan":true,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":24000,"rerank_vectors":null,"soar_lambda":1.0}
mkt-n8000-ccfalse-fs16-soar: {"boundary_epsilon":0.05,"centroid_compression":false,"centroid_fastscan":null,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":8000,"rerank_vectors":null,"soar_lambda":1.0}
mkt-n8000-cctrue-fs16-soar: {"boundary_epsilon":0.05,"centroid_compression":true,"fan_out":null,"fastscan":true,"fastscan_bits":null,"nlist":8000,"rerank_vectors":null,"soar_lambda":1.0}

PostgreSQL settings

effective_cache_size = 384 GB
effective_io_concurrency = 256
huge_pages = try
maintenance_work_mem = 200 GB
max_connections = 100
max_parallel_maintenance_workers = 32
max_parallel_workers = 64
random_page_cost = 1.1
shared_buffers = 128 GB
wal_level = replica
work_mem = 1 GB