Inception42 reports that the post-graph-rag model achieved 10.4% overall performance on a temporal reasoning question, 10.0% on multi-session reasoning, and 10.5% on knowledge update reasoning compared to 71.2% on Zep/Graphiti and 60.2% on full-context GPT-4o.