Cohere-100M — 2026-07-16

r8g.24xlarge · peak recall 0.9965 @ 28 QPS

“100M Cohere on Graviton”

QPS at recall

Recall	Meerkat	pgvector HNSW (halfvec)	Δ vs best competitor
0.90	558	356	+57%
0.95	259	176	+47%
0.99	61	— (max r=0.986)	—

Latency at recall

	Meerkat	pgvector HNSW (halfvec)	Δ vs best competitor
@ recall 0.90
p50	1.76 ms	2.84 ms	-38%
p95	2.24 ms	3.83 ms	-42%
p99	2.73 ms	4.62 ms	-41%
@ recall 0.95
p50	3.78 ms	5.89 ms	-36%
p95	4.77 ms	7.28 ms	-34%
p99	5.96 ms	8.03 ms	-26%
@ recall 0.99
p50	16.17 ms	—	—
p95	19.94 ms	—	—
p99	26.02 ms	—	—

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-defaults · pgvector HNSW (halfvec) @0.90: hnsw-m24-halfvec · Meerkat @0.95: mkt-defaults · pgvector HNSW (halfvec) @0.95: hnsw-m24-halfvec · Meerkat @0.99: mkt-defaults

Build comparison

	Meerkat	pgvector HNSW (halfvec)	Δ vs best competitor
Build time	5699 s	17308 s	-67%
Index size	24.9 GB	179.2 GB	-86%

Build stats are for the configuration backing each column's best operating point: Meerkat = mkt-defaults, prewarmed · pgvector HNSW (halfvec) = hnsw-m24-halfvec, prewarmed.

Plots

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

Host

Instance type	r8g.24xlarge
Hostname	ip-172-31-37-173
CPU	unknown (96 vCPU)
Memory	742 GB
OS	Ubuntu 26.04 LTS
Kernel	7.0.0-1008-aws

Versions

rekall	0.1.0 (764db1b / main)
PostgreSQL	PostgreSQL 18.4 (Ubuntu 18.4-1.pgdg26.04+1) on aarch64-unknown-linux-gnu, compiled by gcc (Ubuntu 15.2.0-16ubuntu1) 15.2.0, 64-bit
meerkat	0.1.0 (b50abf1)
pg_prewarm	1.2
vector	0.8.5

Search

k	10
runs	3
warmup	500
num_queries	10000
S3 archive	s3://rekall-bench/runs/cohere-100m/2026-07-16/r8g.24xlarge-122500/

Index builds

Index	Type	Build time	Size	Workers
hnsw-m24-halfvec	pgvector-hnsw	17307.6s	179.2 GB	32
mkt-defaults	meerkat	5698.7s	24.9 GB	32

Index parameters

hnsw-m24-halfvec: {"ef_construction":200,"encoding":"halfvec","m":24}
mkt-defaults: {"boundary_epsilon":null,"centroid_compression":null,"centroid_fastscan":null,"fan_out":null,"fastscan":null,"fastscan_bits":null,"nlist":null,"rerank_vectors":null,"soar_lambda":null}

PostgreSQL settings

effective_cache_size = 576 GB
effective_io_concurrency = 256
huge_pages = try
maintenance_work_mem = 307 GB
max_connections = 100
max_parallel_maintenance_workers = 32
max_parallel_workers = 96
random_page_cost = 1.1
shared_buffers = 192 GB
wal_level = replica
work_mem = 1 GB