Cohere-100M — 2026-09-28

r8i.8xlarge · peak recall 0.9966 @ 26 QPS

“Testing for regressions before public release”

QPS at recall

Recall	mkt-defaults
0.90	533
0.95	254
0.99	57

Latency at recall

	mkt-defaults
@ recall 0.90
p50	1.83 ms
p95	2.40 ms
p99	2.85 ms
@ recall 0.95
p50	3.83 ms
p95	5.02 ms
p99	5.97 ms
@ recall 0.99
p50	16.98 ms
p95	21.70 ms
p99	26.66 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.

Build comparison

	mkt-defaults
Build time	3960 s
Index size	24.9 GB

Plots

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

Host

Instance type	r8i.8xlarge
Hostname	ip-172-31-37-72
CPU	Intel(R) Xeon(R) 6975P-C (32 vCPU)
Memory	248 GB
OS	Ubuntu 26.04.1 LTS
Kernel	7.0.0-1012-aws

Versions

rekall	0.1.0 (ddee801 / main)
PostgreSQL	PostgreSQL 18.6 (Ubuntu 18.6-1.pgdg26.04+2) on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 15.2.0-16ubuntu1) 15.2.0, 64-bit
pg_prewarm	1.2
pg_vectorsearch	0.1.0-dev (5fb368b)
vector	0.8.6

Search

k	10
runs	3
warmup	500
num_queries	10000
S3 archive	s3://rekall-bench/runs/cohere-100m/2026-09-28/r8i.8xlarge-221915/

Index builds

Index	Type	Build time	Size	Workers
mkt-defaults	prism	3960.2s	24.9 GB	16

Index parameters

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}

Instance settings (observed)

instance type          r8i.8xlarge
vCPU (observed)        32
memory (observed)      247.7 GiB
cpu model              Intel(R) Xeon(R) 6975P-C
kernel                 7.0.0-1012-aws
data directory         /var/lib/postgresql/18/main
mount point            /
filesystem             ext4
media                  ebs
device                 nvme0n1
device size            2147 GB
ebs volume             vol-0fb8a8ebfe4eb1834
ebs volume type        gp3
provisioned IOPS       16000
provisioned throughput 1000 MB/s
io scheduler           none
nr_requests            63
read_ahead_kb          128

PostgreSQL settings

checkpoint_completion_target = 0.9
cluster_name = 18/main
data_directory = /var/lib/postgresql/18/main
effective_cache_size = 192 GB
effective_io_concurrency = 256
huge_pages = try
jit = off
maintenance_io_concurrency = 256
maintenance_work_mem = 102 GB
max_connections = 100
max_parallel_maintenance_workers = 16
max_parallel_workers = 32
max_wal_size = 32 GB
random_page_cost = 1.1
server_version = 18.6 (Ubuntu 18.6-1.pgdg26.04+2)
shared_buffers = 64 GB
wal_level = replica
work_mem = 1 GB