Agent Memory Leaderboard Cycle 2: a fair table for picking a memory layer
Desperate-Ride8765 · reddit · 2026-11-18
Launch posts and demos make picking an agent memory layer a mess — every claim lives in its own setup (answering model, prompts, grading, retrieval depth all differ). A uniformly graded public table is one of the few evidence-based options:
- Agent Memory Leaderboard Cycle 2 results expected mid-November, split into Textual, Coding and Multimodal boards
- Each track has separate Open-source Methods and Commercial Products boards; the answering side is locked for all entries, so score gaps mainly reflect the memory system
- Cycle 1 drew 136 team registrations, 67 memory frameworks completed evaluation (published Aug 12, 2026); top eight within 1.1 points, first vs third just 0.22 apart on Textual Open-source
- Results only hold within their division, version, suite and cycle, per official release notes
Related event: Agent Memory Leaderboard Adds Multimodal Memory Track(2 posts)→
More from coding & agent
- Blueframe: A Browser Canvas for Wireframing UI Structure Before AI Builds It — nima_owji · 2026-11-18
- Neo4j Open-Sources Relate: An Embedded Rust Graph Database with Built-in Agent Memory — JeremyCMorgan · 2026-11-18
- Open-source mail MCP server adds OpenPGP end-to-end encryption, seeks security review — Soft-Lie-434 · 2026-11-18
- Beyond UX: AX and UAX are the new product layers in the agent era — yenkel · 2026-11-18
- Floot Design launches: compare six app versions on one canvas in ChatGPT — CodeByPoonam · 2026-11-18
- Gemini can now call businesses for you to compare quotes and book appointments — GeminiApp · 2026-11-18