chessbench.orgRanks LLMs by their ability to plan, predict and adapt.
MadeByLupu@bylupu
MadeByLupu@bylupu
Results updated 2026-10-26 UTC

GLM 5.3 chess results⁠

Compare GLM 5.3's saved thinking setups, ChessBench ratings, time, cost, and match records.

Full leaderboard

ChessBench Elo ratings compare LLMs within this benchmark only. They are not equivalent to Chess.com, FIDE, or other chess ratings. Beating a 2000-rated LLM here does not establish your rating or predict the same result against a 2000-rated human player.

High #20

1501 ChessBench Elo

Placement complete · 120 rated games

Wins / draws / losses
45 / 17 / 58
Result percentages
38% / 14% / 48%
Time per turn
60.6s
Cost per game
$2.133
Output tokens per game
157,253
Rejected %
1.76%

Max #21

1491 ChessBench Elo

Placement complete · 120 rated games

Wins / draws / losses
47 / 11 / 62
Result percentages
39% / 9% / 52%
Time per turn
60.2s
Cost per game
$1.865
Output tokens per game
138,162
Rejected %
1.61%

How to read this comparison

Each thinking setup has its own rating and completed-game record. Higher thinking settings do not guarantee more wins. The figures above use the same saved data as the leaderboard.

Time includes model requests per accepted move. Cost uses fully priced completed benchmark games. Output includes reported reasoning tokens. Human challenges are separate.

Exact rules and promptsThinking levels and costThis model in MarkdownData and citation guide

Public match records

Page 7 of 12. 238 visible replays. Rating totals include all retained rated results; an archive-hidden result has no link here.

  1. GPT-5.6 Luna (xHigh) vs GLM 5.3 (Max)1/2-1/2

    Insufficient material · 117 half-moves · 2026-09-17 UTC

  2. GPT-5.6 Luna (xHigh) vs GLM 5.3 (High)1/2-1/2

    Draw at 150 total moves · 150 half-moves · 2026-09-18 UTC

  3. Muse Spark 1.3 (xHigh) vs GLM 5.3 (High)0-1

    Checkmate · 94 half-moves · 2026-09-17 UTC

  4. GLM 5.3 (Max) vs GPT-5.6 SOL (Medium)0-1

    Checkmate · 60 half-moves · 2026-09-17 UTC

  5. GLM 5.3 (High) vs Muse Spark 1.3 (Medium)0-1

    Checkmate · 72 half-moves · 2026-09-17 UTC

  6. GPT-5.6 SOL (Medium) vs GLM 5.3 (Max)1-0

    Checkmate · 97 half-moves · 2026-09-17 UTC

  7. Muse Spark 1.3 (Medium) vs GLM 5.3 (Max)0-1

    Checkmate · 46 half-moves · 2026-09-17 UTC

  8. GLM 5.3 (High) vs Muse Spark 1.3 (xHigh)1-0

    Checkmate · 77 half-moves · 2026-09-17 UTC

  9. GLM 5.3 (Max) vs MiniMax M3 (Thinking)1-0

    Checkmate · 121 half-moves · 2026-09-17 UTC

  10. GPT-5.6 SOL (Medium) vs GLM 5.3 (High)1-0

    Checkmate · 47 half-moves · 2026-09-17 UTC

  11. GLM 5.3 (High) vs Muse Spark 1.3 (xHigh)0-1

    Checkmate · 42 half-moves · 2026-09-17 UTC

  12. Muse Spark 1.3 (Medium) vs GLM 5.3 (Max)0-1

    Checkmate · 64 half-moves · 2026-09-17 UTC

  13. GLM 5.3 (Max) vs GPT-5.6 Terra (xHigh)0-1

    Checkmate · 90 half-moves · 2026-09-17 UTC

  14. Muse Spark 1.3 (Medium) vs GLM 5.3 (High)1/2-1/2

    Stalemate · 93 half-moves · 2026-09-17 UTC

  15. GLM 5.3 (Max) vs GPT-5.6 SOL (Medium)0-1

    Checkmate · 60 half-moves · 2026-09-16 UTC

  16. GLM 5.3 (High) vs Muse Spark 1.3 (xHigh)0-1

    Checkmate · 96 half-moves · 2026-09-16 UTC

  17. GLM 5.3 (Max) vs Muse Spark 1.3 (Medium)1/2-1/2

    Threefold repetition · 124 half-moves · 2026-09-16 UTC

  18. GPT-5.6 Luna (xHigh) vs GLM 5.3 (High)1-0

    Checkmate · 99 half-moves · 2026-09-16 UTC

  19. Muse Spark 1.3 (xHigh) vs GLM 5.3 (High)1/2-1/2

    Draw at 150 total moves · 150 half-moves · 2026-09-16 UTC

  20. GPT-5.6 SOL (Medium) vs GLM 5.3 (Max)1/2-1/2

    Threefold repetition · 145 half-moves · 2026-09-16 UTC

Cite these results

ChessBench. GLM 5.3 chess results. https://chessbench.org/models/glm-5.3. Retrieved 2026-10-28. Include the model, thinking level, protocol, and sample size when citing results.

Public data is subject to the current terms. Model names and logos belong to their owners.