Ceibo is a UCI-compatible chess engine written in C++. It can be used to play chess or analyze games through any UCI-compatible chess GUI, such as Arena 3.5.1 or other popular chess interfaces.
The latest stable release is Ceibo 1.10 and it is available free of charge from the Downloads page.
Ceibo is an independently developed chess engine focused on strong search, modern neural network evaluation, and continuous improvement. The project has evolved over time through experimentation with different techniques in computer chess, including search algorithms, evaluation methods, and neural network training.
Universal Chess Interface (UCI) protocol support
Alpha-Beta Search
Principal Variation Search (PVS)
Iterative Deepening
Transposition Tables (up to 8 GB)
Syzygy Endgame Tablebase support
Null Move Pruning
Futility Pruning
Reverse Futility Pruning
Delta Pruning
Late Move Reductions (LMR)
Search Extensions
Quiescence Search
Check Extensions
Static Exchange Evaluation (SEE)
MVV/LVA
History Heuristics
Killer Move Heuristics
Ceibo 1.10 introduces a completely redesigned evaluation system based on NNUE:
Custom NNUE evaluation
King-centric HalfKP feature set
Efficient incremental evaluation updates
Neural network size: approximately 20 MB
Parameter count: approximately 5 million
Network trained using a custom C++/LibTorch training framework
Training data generated from Ceibo self-play games
Written entirely in C++
Custom neural network training pipeline
Continuous engine tuning and optimization
Designed for modern multi-core processors
Ceibo is a long-term hobby project developed independently and released freely for the computer chess community.