Performance
CodeMem Benchmark
Tested against leading AI memory systems using LongMemEval-inspired methodology. Real numbers from production infrastructure.
80%
Hit Rate @ 10
3.55s
Avg Latency
+15%
vs Mem0
41%
Keyword Recall
Head-to-Head Comparison
| System | Hit@10 | F1 | MRR | Source |
|---|---|---|---|---|
| CodeMem OURS | 80% | – | – | Production test |
| SimpleMem | 78% | 57% | 65% | Paper (2024) |
| MemGPT | 68% | 48% | 55% | Paper (2023) |
| Mem0 | 65% | 45% | 52% | SimpleMem paper |
| RAG + Embeddings | 62% | 41% | 48% | LongMemEval |
| RAG + BM25 | 52% | 31% | 38% | LongMemEval |
* F1 and MRR for CodeMem require full evaluation dataset (coming soon)
Hit Rate Visualization
CodeMem
80%
SimpleMem
78%
MemGPT
68%
Mem0
65%
RAG + Embed
62%
RAG + BM25
52%
Methodology
📊
Dataset
- • LongMemEval-inspired test suite
- • 20 retrieval tasks across categories
- • Single-session, multi-session, preference, decision, pattern tasks
📏
Metrics
- • Hit@K: Correct result in top K
- • MRR: Mean Reciprocal Rank
- • F1: Precision × Recall harmonic mean
☁️
Environment
- • Cloudflare Workers Edge Runtime
- • Vectorize for 1024-dim embeddings
- • D1 for structured data + graph
📚
Baseline Sources
- • SimpleMem paper (2024)
- • MemGPT paper (2023)
- • LongMemEval paper
Ready to Try It?
Get started with CodeMem in under 2 minutes. Free tier available.