# Research & Product Updates — Maximem field notes

> Maximem's AI memory research and product updates: LongMemEval and LoCoMo benchmark results, architecture deep-dives on Maximem Synap, launches and release notes.

Maximem Synap scores 92% on LongMemEval and 93.2% on LoCoMo, and the posts here explain what those numbers measure, how we ran them and where they fall short. Benchmarks for memory are young, so we publish the configuration alongside the score.

Alongside the benchmarks sit retrieval experiments run on real corpora, deep-dives into how Synap stores and retrieves context across its vector, graph and file stores, and the product news that follows from that work: launches and release notes, plus guides to getting started with Maximem Synap and Maximem Vity.

## All 7 articles

- [Maximem Synap Updates: Higher Scores, 17 Integrations, and a Live Playground](https://www.maximem.ai/blog/maximem-synap-updates-higher-benchmark-scores-and-more) — May 27, 2026: Synap updates: 92% LongMemEval (up from 90.2%), 93.2% LOCOMO, 17 framework integrations, a browser playground, public pricing, and a free accuracy eval on your own agent.
- [How Synap Works Under the Hood](https://www.maximem.ai/blog/how-maximem-synap-works) — April 11, 2026: We launched Maximem Synap today. Here's a peek into how it is built. 
- [Synap Scores 92% on LongMemEval, 93.2% on LoCoMo: What the Numbers Mean](https://www.maximem.ai/blog/synap-benchmark-results) — April 10, 2026: Synap outperforms existing memory systems by redesigning context management on leading benchmarks; delivering higher accuracy, lower latency, and stable performance at scale through structured, domain-specific architectures.

- [Why We Built Synap](https://www.maximem.ai/blog/why-we-built-synap) — April 10, 2026: AI agents don’t fail from lack of memory, they fail because context doesn’t evolve. This article shows why current approaches break, introduces the Context Management Trilemma, and how Synap enables agents to learn, adapt, and stop forgetting over time.
- [File Search vs Vector Search for RAG: 50,000 Documents, 5,000 Queries, Full Results](https://www.maximem.ai/blog/file-rag-vs-vector-rag) — 2026-01-15: We ran exact-match keyword search against vector search across five datasets, 50,000 documents and 5,000 queries, scored by exact document ID with no LLM judge. Vector search wins on average, and the entire margin comes from one dataset.
- [Introducing Maximem: AI Memory That Actually Works](https://www.maximem.ai/blog/introducing-maximem) — July 25, 2025: Exploring the next generation of AI capabilities and how memory will shape the future of human-AI interaction.
- [How to Get Started with Maximem](https://www.maximem.ai/blog/getting-started-maximem) — July 25, 2025: A step-by-step guide to setting up your AI memory system and maximizing productivity.

Start here: [Benchmark methodology](https://www.maximem.ai/evals) · [How Synap works](https://www.maximem.ai/synap/how-it-works) · [Measuring agent memory](https://www.maximem.ai/measuring-agent-memory) · [Maximem Synap](https://www.maximem.ai/synap)

Other topics: [AI Agent Memory](https://www.maximem.ai/blog/topic/agent-memory) · [Context Engineering](https://www.maximem.ai/blog/topic/context-engineering) · [RAG & Retrieval](https://www.maximem.ai/blog/topic/rag) · [MCP & Agent Protocols](https://www.maximem.ai/blog/topic/mcp) · [Agent Frameworks](https://www.maximem.ai/blog/topic/frameworks) · [Agent Evals & Observability](https://www.maximem.ai/blog/topic/evals) · [Claude Code & Coding Agents](https://www.maximem.ai/blog/topic/coding-agents) · [Voice Agents](https://www.maximem.ai/blog/topic/voice-agents) · [LLM Cost & Production](https://www.maximem.ai/blog/topic/production)

---

Source: [https://www.maximem.ai/blog/topic/research](https://www.maximem.ai/blog/topic/research)