Source support
Prefer primary specifications, announcements and benchmark documentation. Keep capture dates and testing conditions attached to observations. Distinguish vendor-run results from independent evaluations and from our own measurements.
Examples and testing
Label illustrative arithmetic and offline fixtures. An offline test does not establish live API compatibility, model quality, OCR accuracy or production reliability. Report what ran, what did not, and unresolved conditions.
Writing and review
We use AI-assisted drafting and human editorial direction. Separate writing and review passes check support, originality and usefulness. Length alone is not evidence of quality or a reason for a search engine to rank a page.
Corrections
Report a disputed claim to hello@llmscorebook.com with the page URL and evidence. Material corrections should explain the changed claim and its source. Preserve dated observations when later measurements use another setup.
Commercial independence
Sponsorship is labelled. Payment does not determine a benchmark score, reference rank or editorial conclusion. Affiliate or paid links, if introduced, should be disclosed beside the relevant recommendation.