← All systems
MIRIX
OrganizationMIRIX AI (UCSD/NYU founders)
Founded2025
Fundingundisclosed
CodeApache-2.0 · 3.6k★
Multi-agent memory with six typed stores (core/episodic/semantic/procedural/resource/vault).
Every number on file
| claimed | benchmark | config | runner | status |
|---|---|---|---|---|
| 85.38 | LoCoMo | gpt-4.1-mini — full-context row scores 87.52, above it | self (paper) | SELF-REPORTED |
| 59.50% | ScreenshotVQA | own benchmark | self (paper) | SELF-REPORTED |
Rows are not comparable — configs differ. A row graduates to a verdict only through a registered re-run. The protocol →