Created in January 20, 2026
2026
We released a new sparse attention model for long context, see preprint.