Benchmarks

/project-overview/benchmarks

Ruflo publishes a SOTA matrix comparing v3.8.0 against LangGraph / AutoGen / CrewAI on darwin-arm64 + linux-x64. On these workloads Ruflo wins cold start, single turn, and RSS by 1.3×–1953×.

Resource	Link
Headline matrix (gist)	v3.8.0 SOTA matrix
Workload spec	`sota-workload-spec.md`
Progress log	`SOTA-PROGRESS.md`
Raw matrix — darwin	`sota-matrix.json`
Raw matrix — linux	`sota-matrix-linux.json`

For the intelligence-system numbers specifically (HNSW vs brute force, recall@10, crossover N), see the intelligence audit and scripts/benchmark-intelligence.mjs.

Federation example

Verification