Context memory benchmarks
Measured on the memory benchmarks the field actually uses.
MatrixArk's context engine is built on TemporalStore. We benchmark it head-to-head against the leading memory systems on LOCOMO and LongMemEval using one shared open-source harness — the same reader model and embeddings for every system, scored by an LLM judge — so the comparison is about the memory layer, not the model around it.
Head-to-head
TemporalStore matches or beats OpenViking, led by long-horizon memory.
OpenViking / VikingMem report beating the other memory systems in their own papers and pages, so we benchmark against it as the bar to clear. Same harness, same reader, same embeddings; LLM-judged.
| Benchmark | Metric | TemporalStore | OpenViking |
|---|---|---|---|
| Overall | Answer accuracy (LLM judge) | 42% | 34.7% |
| LongMemEval | Answer accuracy | 38% | 16% |
| LOCOMO | Answer accuracy | 44% | 44% |
| LongMemEval | Retrieval hit@k | 0.98 | 0.81 |
| Context packs | Prompt tokens at equal quality | Large reduction vs replaying full history | |
The gap is widest on LongMemEval — long-horizon, multi-session questions — exactly where time-valid, replayable memory with stale-context blocking pays off. On shorter LOCOMO conversations the two are even.
The landscape
How the context approaches differ.
| System | Approach | Temporal & replay | One service |
|---|---|---|---|
| MatrixArk / TemporalStore | Append-structured temporal store; Context model | Native event-time + replay | Yes — memory, retrieval, counters, replay |
| OpenViking / VikingMem | Event/entity memory, L0/L1/L2 layers | Temporal layers | Memory-focused |
| Mem0 | Add/search API, scoped memories | Limited | Needs a vector store |
| Zep / Graphiti | Bi-temporal knowledge graph | Bi-temporal graph | Graph + retrieval |
| MemOS / Memori | OS/filesystem-style memory | File-like | Memory-focused |
Method
Fair by construction.
Both systems run behind the same open-source harness: identical reader model and embedding model, the same LOCOMO (multi-session conversation QA) and LongMemEval (long-horizon memory) datasets, and an LLM judge applied uniformly. Only the memory layer differs, so the results isolate the contribution of the store — ingestion, retrieval, freshness, and context-pack assembly.
The enterprise story adds the parts a benchmark cannot show on a laptop: shared-storage throughput under concurrency, five-nines availability, and replayable audit — all on the same engine.
Read more