Ceres Chess Engine

Ceres Chess Engine is an open-source, UCI-compatible chess engine built in C# on the .NET framework. Developed by dje-dev, Ceres uses AlphaZero-style Monte Carlo Tree Search (MCTS) combined with deep neural networks, making it both a competitive playing engine and a research platform for exploring neural network-based chess AI.

The project spans approximately 80,000 lines of code across 500 source files and includes advanced capabilities like Non-Linear Attention (NLA) neural network augmentation, transposition detection to avoid redundant evaluations, and support for multiple GPU inference backends including CUDA and TensorRT. It also ships with integrated tools for position analysis, suite testing, and tournament management — useful for researchers and engine developers alike.

Ceres has earned its place on the competitive stage, reaching the top leagues of TCEC (Top Chess Engine Championship). Highlights include 1st place in both TCEC Swiss 7's Entrance League and League 2 in late 2024, plus top-four finishes in TCEC Cup 14. It is free and open-source under the GNU GPL v3, making it accessible to anyone interested in cutting-edge chess AI research.