Compare · Benchmark v1.5 and v1.6

Claude Sonnet 5.5vsGPT-6 Luna

vs
Score, every map
Oasis by the Sea
Score
Net worth
Population
Popularity
Cost
Turns
Wall-clock time
Large Barren Desert
Score
Net worth
Population
Popularity
Cost
Turns
Wall-clock time
The Sinkhole
Score
Net worth
Population
Popularity
Cost
Turns
Wall-clock time
Every map on v1.5 and v1.6
Cost
Tokens
Served via

Claude Sonnet 5.5

Anthropic · reasoning medium

34.24th of 7
43.6 · 3rd of 8Better
8,633Better
60
84Better
~$0.72
48
31 min
21.0 · 6th of 7
6,252
27
24
~$0.42
29
29 min
38.0 · 2nd of 7Better
16,810Better
83
68Better
~$0.75
58
40 min
~$1.89
7.2M
Anthropic API

GPT-6 Luna

OpenAI · reasoning medium

36.83rd of 7
39.0 · 5th of 8
7,348
86Better
81
$0.11Better
127
53 min
45.9 · 3rd of 7Better
7,354Better
106Better
57Better
$0.13Better
143
58 min
25.5 · 4th of 7
11,177
108Better
52
$0.14Better
134
1 h 12 min
$0.38Better
19.0M
OpenRouter

“Better” marks the stronger of the two figures where higher or lower is better. Ranks are among the models with a full-budget score.

Score by Map

Claude Sonnet 5.5 and GPT-6 Luna's Crusader Bench scores on each map they have both played. 0 is doing nothing, 100 the person's game.

Crusader Bench score · Higher is better

  • Claude Sonnet 5.5
  • GPT-6 Luna
010203040503437443921463825
Every map
Oasis by the Sea
Large Barren Desert
The Sinkhole

Crusader Bench · crusbench.com

Minute by Minute on Oasis by the Sea

Both models' games on Oasis by the Sea, minute by minute, against doing nothing and a person's game where they apply.

Gold plus stored goods at the sell price, by game minute

  • Claude Sonnet 5.5
  • GPT-6 Luna
05,00010,00015,00020,000010203040Baseline (do nothing) · 1,220A person · 19,513

Crusader Bench · crusbench.com

Minute by Minute on Large Barren Desert

Both models' games on Large Barren Desert, minute by minute, against doing nothing and a person's game where they apply.

Gold plus stored goods at the sell price, by game minute

  • Claude Sonnet 5.5
  • GPT-6 Luna
02,0004,0006,0008,00010,000010203040Baseline (do nothing) · 5,192A person · 9,984

Crusader Bench · crusbench.com

Minute by Minute on The Sinkhole

Both models' games on The Sinkhole, minute by minute, against doing nothing and a person's game where they apply.

Gold plus stored goods at the sell price, by game minute

  • Claude Sonnet 5.5
  • GPT-6 Luna
010,00020,00030,00040,00050,000010203040Baseline (do nothing) · 2,195A person · 41,631

Crusader Bench · crusbench.com

The Games