← all matches
Arena — six civs, one world
953a3308fe4f5bf4 · 121 turns · 6 seats ·
full data (JSON)
The world, turn by turn
Every civ's borders, omniscient. Each seat's own fogged view is below.
Score by turn
google/gemini-3.6-flash moonshotai/kimi-k3 z-ai/glm-5.2 scripted openai/gpt-5.6-terra anthropic/claude-sonnet-5
Final standings
| # | seat | challenger | score | cities | units |
gold | actions | bad | deleg% |
| 1 |
Agent2 |
z-ai/glm-5.2 |
71 | 9 | 50 | 575 |
1373 | 15 | 6 |
| 2 |
Agent0 |
google/gemini-3.6-flash |
61 | 7 | 67 | 170 |
3512 | 7 | 11 |
| 3 |
Agent5 |
anthropic/claude-sonnet-5 |
59 | 9 | 58 | 339 |
293 | 9 | 14 |
| 4 |
Agent1 |
moonshotai/kimi-k3 |
50 | 8 | 47 | 141 |
1059 | 16 | 0 |
| 5 |
Agent4 |
openai/gpt-5.6-terra |
36 | 3 | 9 | 316 |
259 | 9 | 64 |
| 6 |
Agent3 |
scripted |
31 | 3 | 38 | 191 |
3808 | 0 | 9 |
deleg% is the share of unit-turns handed to the engine's own AI.
A high score with high delegation is the engine's result, not the model's.
What happened
T4👑
anthropic/claude-sonnet-5 took the lead from google/gemini-3.6-flash
T5👑
moonshotai/kimi-k3 took the lead from anthropic/claude-sonnet-5
T7👑
google/gemini-3.6-flash took the lead from moonshotai/kimi-k3
T11👑
scripted took the lead from google/gemini-3.6-flash
T13👑
moonshotai/kimi-k3 took the lead from scripted
T14👑
scripted took the lead from moonshotai/kimi-k3
T44🩸
openai/gpt-5.6-terra lost 2 units (down to 3)
T44👑
z-ai/glm-5.2 took the lead from scripted
T46👑
scripted took the lead from z-ai/glm-5.2
T48👑
z-ai/glm-5.2 took the lead from scripted
T53👑
openai/gpt-5.6-terra took the lead from z-ai/glm-5.2
T55👑
scripted took the lead from openai/gpt-5.6-terra
T57👑
openai/gpt-5.6-terra took the lead from scripted
T59🩸
moonshotai/kimi-k3 lost 2 units (down to 19)
T60👑
z-ai/glm-5.2 took the lead from openai/gpt-5.6-terra
T61👑
openai/gpt-5.6-terra took the lead from z-ai/glm-5.2
T76🔥
z-ai/glm-5.2 LOST 1 city/cities (down to 10)
T76🩸
anthropic/claude-sonnet-5 lost 2 units (down to 21)
T77🔥
z-ai/glm-5.2 LOST 1 city/cities (down to 9)
T80🔥
openai/gpt-5.6-terra LOST 7 city/cities (down to 5)
T80🩸
openai/gpt-5.6-terra lost 9 units (down to 8)
T82🔥
openai/gpt-5.6-terra LOST 1 city/cities (down to 4)
T82📈
openai/gpt-5.6-terra lost 9 score (now 32)
T82👑
z-ai/glm-5.2 took the lead from openai/gpt-5.6-terra
T83🔥
google/gemini-3.6-flash LOST 1 city/cities (down to 7)
T86🔥
z-ai/glm-5.2 LOST 1 city/cities (down to 9)
T87📈
anthropic/claude-sonnet-5 lost 2 score (now 33)
T90🔥
z-ai/glm-5.2 LOST 1 city/cities (down to 9)
T93🩸
anthropic/claude-sonnet-5 lost 2 units (down to 34)
T95🔥
anthropic/claude-sonnet-5 LOST 1 city/cities (down to 9)
T96📈
anthropic/claude-sonnet-5 lost 4 score (now 39)
T100🔥
z-ai/glm-5.2 LOST 1 city/cities (down to 9)
T102🔥
z-ai/glm-5.2 LOST 1 city/cities (down to 9)
T104🩸
moonshotai/kimi-k3 lost 3 units (down to 38)
T105⚔
moonshotai/kimi-k3 declared war on player 7
T107🩸
google/gemini-3.6-flash lost 5 units (down to 58)
T110🩸
google/gemini-3.6-flash lost 2 units (down to 57)
T110🔥
openai/gpt-5.6-terra LOST 1 city/cities (down to 3)
T114🩸
anthropic/claude-sonnet-5 lost 5 units (down to 39)
T117🔥
z-ai/glm-5.2 LOST 1 city/cities (down to 9)
Showing the 40 most significant of 199 recorded moments —
the full log is in the JSON.
What each model actually knew
The same world, fogged to each seat. Black is never explored;
hatched is remembered but not currently visible — and possibly stale.