Reference

Fine-tuned models

GNO ships a fine-tuned local expansion model trained on Apple Silicon with MLX LoRA.

GNO publishes a fine-tuned retrieval expansion model —gno-expansion-slim-retrieval-v1 — trained locally on Apple Silicon with MLX LoRA. It’s the default on fresh installs and benchmarks meaningfully above the base slim preset.

What this is

The expansion model rewrites or augments difficult queries before retrieval. It is not the model that stores your documents or answers every question; it is a small local helper that improves recall when your query and your notes use different words.

The numbers

• nDCG@10 0.925 on the canonical benchmark corpus
• Ask Recall@5 0.875 on the ask-style answer benchmark
• Shipped via Hugging Face, pulled on first run

Install it

gno models use slim-tuned
gno models pull

Train your own

The training harness lives in the main GNO repo under research/. It uses MLX LoRA for local fine-tuning and can export GGUF artifacts you can install via the Hugging-Face-backed custom preset format.

Benchmarks

Before promoting a model change, run the benchmark suite. Use gno bench for a public CLI check against your own fixture, and see the benchmarks feature page for the fixture workflow and result metrics.

gno bench docs/examples/bench-fixture.json
gno bench fixture.json --json