Skip to content
AI News HubLIVE
More
Summary only

#1 on the leading AI memory benchmark using a smaller, cheaper model

Exabase's M-1 memory engine achieves state-of-the-art on LongMemEval with 96.4% accuracy using Gemini 3 Flash, a smaller model, without prompt engineering, demonstrating efficient retrieval architecture.

SourceHacker News AIAuthor: johnnymakes
#1 on the leading AI memory benchmark using a smaller, cheaper model
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions

Summary only

Under the source content permission, this page provides an editorial summary, key points, and a link to the original article.

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • M-1 achieves 96.4% accuracy at top-50 retrieval depth using a smaller model.
  • Three-phase retrieval pipeline: scoring, query decomposition, and re-ranking.
  • Single-session tasks near perfect; multi-session and temporal reasoning show strong results.
  • Performance peaks at moderate depth, indicating effective ranking rather than reliance on large context.

Highlights and analysis are generated automatically and may contain errors. Check the original source.

More recent stories on this topic · Models