Benchmark Almanac

CivBench

GLM-5.3 1,634 leads since Oct 2026 Not in model cards

Models play full games of Civilization V with the Vox Populi mod against each other on matched maps and seeds. Pairwise skill ratings rank long-term strategy.

Also tracks diplomacy, strategy choices and win prediction quality. The Vox Deorum repo has about 110 stars. Release month is the first Reddit post found.

Made by Vox Deorum Released Oct 2026 Video games All tasks public Last model added Sep 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Ranked by matches Models play against each other, and the results become ratings.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
vox-deorum/vox-deorum 107 stars Leaderboard vox-deorum.github.io
23
Entries on its own leaderboard, 18 models we track

GLM-5.3 leads with 1,634

Best score per entry from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

23 entries. Scroll sideways for the rest. Bars start at 1,300.

  1. 1,634 GLM-5.3
  2. 1,602 GPT-6.1 Sol
  3. 1,593 Claude Opus 5.5
  4. 1,573 Kimi-K2.7
  5. 1,557 GPT-6 Astra
  6. 1,537 DeepSeek V4.1 Flash
  7. 1,535 GLM-5.2
  8. 1,516 Qwen 3.8 27B
  9. 1,510 GPT-6 Luna
  10. 1,504 Kimi K2.6
  11. 1,499 Qwen-3.8-Flash-Next
  12. 1,497 Claude Sonnet 5.5
  13. 1,482 GLM-5.1
  14. 1,477 DeepSeek-V4-Flash
  15. 1,470 MiniMax-M2.7
  16. 1,468 GLM-5.3-Flash
  17. 1,460 Qwen-3.5
  18. 1,458 Gemma-4
  19. 1,448 MiniMax-M3
  20. 1,441 Nemotron 3 Super
  21. 1,422 gpt-oss-120b
  22. 1,391 Qwen3.6 27B
  23. 1,350 Jev!

Bars start at 1,300.

  1. 1 GLM-5.3 1,634
  2. 2 GPT-6.1 Sol 1,602
  3. 3 Claude Opus 5.5 1,593
  4. 4 Kimi-K2.7 1,573
  5. 5 GPT-6 Astra 1,557
  6. 6 DeepSeek V4.1 Flash 1,537
  7. 7 GLM-5.2 1,535
  8. 8 Qwen 3.8 27B 1,516
  9. 9 GPT-6 Luna 1,510
  10. 10 Kimi K2.6 1,504
  11. 11 Qwen-3.8-Flash-Next 1,499
  12. 12 Claude Sonnet 5.5 1,497
  13. 13 GLM-5.1 1,482
  14. 14 DeepSeek-V4-Flash 1,477
  15. 15 MiniMax-M2.7 1,470
  16. 16 GLM-5.3-Flash 1,468
  17. 17 Qwen-3.5 1,460
  18. 18 Gemma-4 1,458
  19. 19 MiniMax-M3 1,448
  20. 20 Nemotron 3 Super 1,441
  21. 21 gpt-oss-120b 1,422
  22. 22 Qwen3.6 27B 1,391
  23. 23 Jev! 1,350
#EntryScoreSettingDateSource
1 GLM-5.3 1,634 every 5 turns Aug 2026 Official leaderboard
2 GPT-6.1 Sol 1,602 every 5 turns Sep 2026 Official leaderboard
3 Claude Opus 5.5 1,593 every 5 turns Sep 2026 Official leaderboard
4 Kimi-K2.7 1,573 every turn Official leaderboard
5 GPT-6 Astra 1,557 every 5 turns Sep 2026 Official leaderboard
6 DeepSeek V4.1 Flash 1,537 every turn Sep 2026 Official leaderboard
7 GLM-5.2 1,535 every turn Jun 2026 Official leaderboard
8 Qwen 3.8 27B 1,516 every turn Aug 2026 Official leaderboard
9 GPT-6 Luna 1,510 every 5 turns Sep 2026 Official leaderboard
10 Kimi K2.6 1,504 every turn Apr 2026 Official leaderboard
11 Qwen-3.8-Flash-Next 1,499 every 5 turns Official leaderboard
12 Claude Sonnet 5.5 1,497 every 5 turns Sep 2026 Official leaderboard
13 GLM-5.1 1,482 every turn Apr 2026 Official leaderboard
14 DeepSeek-V4-Flash 1,477 every 5 turns Apr 2026 Official leaderboard
15 MiniMax-M2.7 1,470 every turn Mar 2026 Official leaderboard
16 GLM-5.3-Flash 1,468 every turn Aug 2026 Official leaderboard
17 Qwen-3.5 1,460 every turn Official leaderboard
18 Gemma-4 1,458 every 5 turns Official leaderboard
19 MiniMax-M3 1,448 every 5 turns Jun 2026 Official leaderboard
20 Nemotron 3 Super 1,441 every turn Mar 2026 Official leaderboard
21 gpt-oss-120b 1,422 every turn Aug 2025 Official leaderboard
22 Qwen3.6 27B 1,391 every turn Apr 2026 Official leaderboard
23 Jev! 1,350 every turn Official leaderboard

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Z.ai
Own leaderboard 1 leader
GLM-5.3 1,634

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
1,300 1,400 1,500 1,600 1,700 2026 1,300 1,400 1,500 1,600 1,700 2026

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Games and long tasks

All 68 in Games and long tasks
  • ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6 Astra 100%
  • MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks. Seed2.0 Pro 49%
  • Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Scores come from in-game progress or outcomes per title. Nemotron 3 Nano 4B 22.9
  • GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. Gemini 3 Flash 41.9%
  • Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%
  • TextQuests Only in system cards Agents play 25 classic interactive fiction games. Measures game progress and harmful actions per 100 steps. Muse Spark 2.40