Platform changelog
What changed.
Changes to how Najd presents and explains its evaluations. Published platform updates; new model evaluations are announced separately.
Najd Arena is live
- Live at najdarena.com with database-backed model metadata and results.
- Responsive navigation, touch-friendly controls, and scrolling detail tables.
- Publication controls require both organization and Najd approval.
- Separate home, model catalog, release browser, and model recommender.
- One setting by default, with expandable thinking-level comparisons and shareable execution settings.
- Task-based recommendations with explicit evidence gaps.
- Configuration-specific task scores, searchable releases, and provider logos.
- Answer-quality breakdowns, CSV export, and exploratory timing under Inference.
No new model evaluation or dataset release is announced by these interface changes.