Open public goods
The four public goods
We're publishing the evaluation infrastructure alongside the model — openly, so any Canadian AI builder, evaluator, or procurer can reference, reproduce, and audit the same baseline. The standard is the point. Links go live as each artifact lands.
Canadian Bilingual Legal Corpus
The open dataset flash-1-mini is fine-tuned on, with full provenance documentation. Bilingual English and French, Canadian legal context.
Coming soonCBLRE Evaluation Suite
The Canadian Bilingual Legal & Regulatory Evaluation — six tracks, bilingual ground truth, reproducible scoring. In preview, pending subject-matter-expert validation.
Coming soonCanadian AI Evaluation Methodology
How to evaluate AI for Canadian regulated workflows — the framework behind the CBLRE tracks.
Coming soonModel Benchmarking Methodology
How we measured what we measured — the reproducibility protocol that makes every published number checkable.
Coming soonMaintained, versioned, and updated by the SimpleDirect team. Reference them in RFPs and procurement scoring; cite them in academic work.
See the model these standards measured
Go to flash-1-mini