Key Info
Tencent's Hy4 model ranked #2 among all models on the ChessQA benchmark, correctly solving 34 of 35 chess puzzles by reasoning carefully instead of answering quickly.
Highlights
- Hy4 placed second across all evaluated models on ChessQA
- Solved 34 of 35 chess puzzles correctly
- The thread argues Hy4 compares well to frontier models as a chess puzzle annotator
- Performance attributed to deliberate reasoning rather than fast, blitz-style answers