Recall vs QPS (Pareto)
“Graviton: 8xlarge downsized from 24xlarge”
cohere-100m | 2026-07-17 | r8g.8xlarge, unknown, 32 vCPU, 247 GB
PG 18.4 | shared_buffers=64 GB | work_mem=1 GB | max_parallel_maintenance_workers=31 | meerkat 0.1.0 (b50abf1), pgvector 0.8.5
All
None