Benchmark Hub
CI platform suite (graph, Sentinel, streaming, ontology) plus interactive browser probes — VibeScript Present purpose clocks, graph, geometry/renderer as their own claim planes.
CI Platform Suite
Live measurements from qualia-cli bench --suite full committed as llm_benchmark_results.json. Includes graph primitives, Sentinel governance, lazy streaming, and WordNet/SHACL highlights.
Graph Operation Benchmarks
Core semantic graph operations — point lookup, two-hop traversal, predicate filtering, and ingestion performance.
execute_ntriples_query + compile_query_to_json timed in this tab. Multi-engine CI suite lives under the Platform tab.
Live RDF graph (queried by execute_ntriples_query) - drag nodes / scroll to zoom
Relationship Explorer
Latency Distribution
Throughput Over Time
VibeScript Present — purpose clocks
checked · compile · run ×2Soft-rise Present clocks: checked cell, CompiledCell.compile, and run() ×2 on the same handle.
Binding footnote — CompiledCell.run
Computational Geometry & Renderer Scene Lab
Bowyer-Watson · Monotone Chain · BVH · Webizen SDFRender a dense, deterministic scene first, then inspect the corresponding browser geometry model. The canvas draws every selected primitive (up to 3,000); its interactive Delaunay preview deliberately uses a bounded subset so it remains responsive. This is a browser Canvas scene preview, not evidence of native GPU performance.
Dense Geometry Scene Preview Every selected primitive is drawn · click to add a solver point
Measured Browser Model Throughput
Cross-Size Measurements
Scientific Computing Primitives
Native scientific primitives integrated with the semantic graph — thermodynamics, ODE solvers, bioinformatics, and clinical risk calculations.
Performance by Domain
Precision Comparison
Spatial Mathematics Benchmarking
Geometric algebra, spatial indexing, and topological operations — benchmarking spatial query performance.
Interactive 3D object — drag to rotate · scroll to zoom
Spatial Operation Performance
Dimension Scaling
Q42 10D Tensor Verification Boundary
This browser page has no binding to the native 10D tensor kernels. It therefore does not invent throughput, latency, heap, or 48-byte compliance results. Select a proposed workload to record what a native benchmark must execute and verify.
Native Throughput Evidence
Dimension Evidence
SPARQL Query Integration
Benchmark SPARQL query performance over QualiaQuin storage — comparing native SPARQL vs QualiaDB optimized queries.
Live query graph - measured matches
SPARQL Match Explorer
Query Latency Distribution
Complexity vs Performance
Zero-Heap Model & Native Verification Boundary
The zero-heap contract applies to declared Tier-1 Rust hot paths, not to this browser page. A browser cannot inspect the native allocator, prove stack-only execution, or certify the Sentinel budget. Use this panel to inspect the selected verification scope; use the native zero-allocation tests for evidence.
zero_heap_testsBrowser Allocation Evidence
Verification Scope
What must be verified natively
Browser LLM Benchmarks
QualiaDB's native WASM + WebGPU LLM (.q42 AOT, ~5.9 tok/s on SmolLM2-360M) vs WebLLM / Transformers.js. Loaded on demand � the LLM engine and comparison adapters initialise only when you open this tab.
QualiaDB 0.0.41 line · tip branch 0.0.41.1 · VibeScript CompiledCell binding