2026-10-11 16:36 UTC

memory-benchmarks

band: coolmomentum: stable score: 0.079
temperature history

Episodes (2)

IreneAI claims the released open Add/Search evaluation framework compares agent-memory systems without team-selected answer models or evaluation pipelines, potentially separating memory quality from evaluation-setup advantages.
watchingconvergesscott: medium
BrainAPI’s developer claims its externally managed memory layer outperforms Mem0, Zep, and Letta on LoCoMo and BEAM1M, which would make BrainAPI a competitive backend for persistent agent memory if the reported results hold.
expiredknownscott: low