MemPalace stores conversations verbatim and retrieves them at 96.6% accuracy. REM Labs scores 94.6% on LongMemEval (473/500) under the byte-exact upstream GPT-4o judge — and runs 9 consolidation strategies to generate knowledge you never stored.
MemPalace is a high-accuracy verbatim retrieval system. It excels at finding exactly what you said. That's where the similarity ends.
| Feature | REM Labs | MemPalace |
|---|---|---|
|
Persistent memory
|
✓ | ✓ |
|
Benchmark accuracy
LongMemEval (ICLR 2025)
|
✓ 94.6% (473/500) | 96.6% |
|
Overnight synthesis (Dream Engine)
AI consolidates knowledge autonomously
|
✓ 9 strategies | ✗ none |
|
Knowledge compression
Lossy compression + tournament
|
✓ | ✗ verbatim only |
|
Knowledge graph
Auto-extracted entities and links
|
✓ | ✗ none |
|
Insight generation
Autonomous cross-memory synthesis
|
✓ | ✗ none |
|
Consumer UI
Designed for everyday people
|
✓ | ✗ CLI only |
|
Storage efficiency
How storage scales over time
|
✓ compressed | ✗ linear growth |
|
Works with any model
|
✓ | ✓ |
MemPalace finds what you said. REM Labs discovers what it means.
Join REM Labs and let your AI finally remember you — and grow smarter every night.
Get started free