Mastra's Observational Memory tops BEAM long-term memory benchmark at 100K and 1M tokens
darian314 · x · 2026-10-30
The team stress-tested Observational Memory on the BEAM long-term memory benchmark at 100K, 1M and 10M tokens with default settings. It posted the top published scores at 100K (83.3%) and 1M (80.6%), ahead of Exabase (76.9%/75.0%), Hindsight (73.4%/73.9%) and Honcho (63.0%/61.8%); at 10M it scored 66.4%, statistically tied with Exabase's 68.0%. A signal that observational memory scales with chat length.
More from coding & agent
- Jack&Jill Matches Closed-Model Quality with Together AI's Open-Model Post-Training — togethercompute · 2026-10-30
- Free search APIs keep dying for local agents — what are you paying per 1,000 queries? — CO_Oked_COO · 2026-10-30
- Install Codex on a separate SSD via junction to stop I/O contention with Windows — Yamapama · 2026-10-30
- Open-source auto-research agent cracks CUDA kernel challenge's global top 10 — IgorCarron · 2026-10-30
- Peter Yang uses Linear's MCP to run his video pipeline with editors and artists — petergyang · 2026-10-30
- A read-only Grok Bot assistant replays your Mac screen and call history to find lost time — bfrench · 2026-10-30