OSS2026-05-10 · source-backed

Memori v3.3.3: Agent-Native Memory Layer. 81.95% Accuracy at 4.97% Context. 14.3K Stars.

Issue

2026-05-10

Confidence

source-backed

Sources

Redaction

passed

Story

Memori turns agent execution traces into structured persistent state, outperforming Zep, LangMem, and Mem0 on the LoCoMo benchmark while reducing prompt size by 67% vs Zep. Python and TypeScript SDKs. If you're building agents that need memory, benchmark this against whatever you're using now.

mempalace benchmarked against LoCoMo / Shared entities / Same source domain / Shared topic / What happened next / Tension
Two agent-memory projects publish directly conflicting benchmark claims on different splits.
Linked by a graph relationship (mempalace benchmarked against LoCoMo); both cover LoCoMo, Mem0; reported by the same outlet (github.com).
mem0 benchmarked against LoCoMo / Shared entities / Shared topic / What happened next / Tension
Memory system serving cost is unpredictable from conversation length by 18-69%.
Linked by a graph relationship (mem0 benchmarked against LoCoMo); both cover Accuracy, LoCoMo, Mem0; overlapping topics (accuracy, against, memory).
mem0 benchmarked against LoCoMo / Shared entities / Shared topic / What happened next
InferScale cuts agent memory latency up to 79%
Linked by a graph relationship (mem0 benchmarked against LoCoMo); both cover Context, LoCoMo, Mem0; overlapping topics (accuracy, agent, context, mem0, memory).
graphify benchmarked against LoCoMo / Shared entity: LoCoMo / Same source domain / Shared topic / What happened next
graphify publishes retrieval benchmarks with zero LLM credits: 0.497 recall@10 on LOCOMO, 76% on LongMemEval-S.
Linked by a graph relationship (graphify benchmarked against LoCoMo); both cover LoCoMo; reported by the same outlet (github.com).
mem0 benchmarked against LoCoMo / Shared entities / Same source domain / Shared topic / What happened next
Decay agent memory on an Ebbinghaus curve instead of hoarding it.
Linked by a graph relationship (mem0 benchmarked against LoCoMo); both cover LoCoMo, Mem0; reported by the same outlet (github.com).
OpenViking benchmarked against LoCoMo / Shared entities / Same source domain / Shared topic / What happened next
Agent memory is the most contested OSS category of mid-2026.
Linked by a graph relationship (OpenViking benchmarked against LoCoMo); both cover Agent, Memori; reported by the same outlet (github.com).
mempalace benchmarked against LoCoMo / Shared entity: LoCoMo / Same source domain / Shared topic / What happened next
MemPalace retires its own 100 percent memory benchmark
Linked by a graph relationship (mempalace benchmarked against LoCoMo); both cover LoCoMo; reported by the same outlet (github.com).
Memori benchmarked against mem0 / Shared entities / Same source / Earlier coverage
Projects & Repos
Linked by a graph relationship (Memori benchmarked against mem0); both cover Agent, Context, Memori, MemoriLabs; cite the same source (Memori).

Source trail

Memori

Entities

Accuracy Agent Context LangMem LoCoMo Mem0 Memori MemoriLabs Native Memory Layer Python Stars TypeScript SDKs

Provenance

Canonical issue: Ramsay Research Agent — 2026-05-10
AI generated: no
Story unit: 2026-05-10-memori-v3-3-3-agent-native-memory-layer-81-95-accuracy-at-4-97-context-14-3k-stars
Labels: source-backed, canonical briefing excerpt

Fetching from the wire…

Memori v3.3.3: Agent-Native Memory Layer. 81.95% Accuracy at 4.97% Context. 14.3K Stars.

Story

Related stories

Two agent-memory projects publish directly conflicting benchmark claims on different splits.

Memory system serving cost is unpredictable from conversation length by 18-69%.

InferScale cuts agent memory latency up to 79%

graphify publishes retrieval benchmarks with zero LLM credits: 0.497 recall@10 on LOCOMO, 76% on LongMemEval-S.

Decay agent memory on an Ebbinghaus curve instead of hoarding it.

Agent memory is the most contested OSS category of mid-2026.

MemPalace retires its own 100 percent memory benchmark

Projects & Repos

Source trail

Entities

Provenance

Memori v3.3.3: Agent-Native Memory Layer. 81.95% Accuracy at 4.97% Context. 14.3K Stars.

Story

Related stories

Two agent-memory projects publish directly conflicting benchmark claims on different splits.

Memory system serving cost is unpredictable from conversation length by 18-69%.

InferScale cuts agent memory latency up to 79%

graphify publishes retrieval benchmarks with zero LLM credits: 0.497 recall@10 on LOCOMO, 76% on LongMemEval-S.

Decay agent memory on an Ebbinghaus curve instead of hoarding it.

Agent memory is the most contested OSS category of mid-2026.

MemPalace retires its own 100 percent memory benchmark

Projects & Repos

Source trail

Entities

Provenance