LLM Chess

Chess Leaderboard

Reasoning Continuation

Auto

White Player

Player Type:

Provider:

Model:

Custom Model:

Leave empty to use selected model above

API Key:

Temperature: 0.7

Black Player

Player Type:

Provider:

Model:

Custom Model:

Leave empty to use selected model above

API Key:

Temperature: 0.7

Max Retries:

Debug Mode

Turn: White Move: 1 Status: In Progress

White: 39 Black: 39 Equal

Piece Set:

Move History & Analysis

This is a modification of LLM Chess, stripping away reasoning, board state, legal moves, and just working by pure PGN movetext continuation.

More likely to produce invalid moves from the models, particularly on black.

I ran a 16-model tournament and documented the results, which you can view here: dubesor.de/chessbeta/tournament2

Security Note: API keys are stored in local storage only (in your browser) and transmitted only for moves to the respective first-party API endpoints.

Home LLM Benchmark GPT-Chars NOPE Puzzle Chess Guestbook