Benchmarks

Benchmark retrieval models and code embedding candidates against fixed corpora, real GNO code slices, and pinned public OSS slices before changing defaults. GNO's research loops are built around measurable wins, not vibes.

Install GNO Read the docs

Use cases: Evaluating a new embedding or reranker before shipping; Comparing candidates against a stable baseline; Generating per-collection model recommendations with receipts

What it gives you

Regression-first retrieval evaluation
Canonical benchmark corpora for stable comparisons
Real GNO code slice benchmark for product-shaped signal
Public OSS slices for generalization checks
Autonomous candidate search with bounded budgets
Per-collection model recommendations backed by results

Try it yourself

Representative commands and entry points. Full reference lives in the documentation.

bun run eval:hybrid

bun run bench:code-embeddings --candidate bge-m3-incumbent --write

bun run research:embeddings:autonomous:search --dry-run

Keep reading

Related features and docs.

Fine-tuned models

/features/fine-tuned-models

Open

Hybrid search

/features/hybrid-search

Open

How search works

/docs/how-search-works

Open

All features

Back to the feature index

Browse all