← Back to leaderboard  ·  Show all task instances

php/php-src

The PHP Interpreter

40,030 c hard
14,288
Generated Behavioral Tests
17.8%
Best Score
Claude Opus 4.8 (xhigh)

Results by Model

Hover for details · Click a model to open it · Drag to zoom the x-axis · Shift-drag to pan · The line marks the Pareto frontier (best score per cost)

Per-Model Results

24 runs
# Model Score Cost Calls
1 Claude Opus 4.8 (xhigh) 17.8% $27.39 167 trace →
2 GPT-5.6 Sol (xhigh) 16.1% $6.78 34 trace →
3 Muse Spark 1.3 (max) 15.0% $7.32 338 trace →
4 GPT 5.5 (xhigh) 13.2% $12.89 115 trace →
5 Gemini 3.7 Flash 12.3% $1.00 57 trace →
6 GPT-5.6 Sol 8.6% $1.05 13 trace →
7 GPT 5.5 (high) 8.0% $3.23 36 trace →
8 Muse Spark 1.2 (xhigh) 5.3% $2.91 252 trace →
9 Claude Haiku 4.5 4.8% $0.52 119 trace →
10 Claude Opus 4.7 (xhigh) 3.2% $3.26 73 trace →
11 GPT 5.5 3.1% $1.47 16 trace →
12 GPT 5.4 2.2% $0.24 7 trace →
13 Gemini 3.1 Pro 1.7% $1.33 109 trace →
14 GPT 5.4 mini 1.1% $0.03 13 trace →
15 Muse Spark 1.3 (xhigh) 1.0% $1.11 187 trace →
16 GPT 5 mini 0.9% $0.04 14 trace →
17 Claude Opus 4.7 0.9% $0.25 14 trace →
18 Gemini 3 Flash 0.6% $0.22 65 trace →
19 Gemini 3.6 Flash 0.3% $3.46 132 trace →
20 Claude Opus 5 (xhigh) 0.0% $1.41 24 trace →
21 GLM-5.2 0.0% $0.39 31 trace →
22 Gemini 3.5 Flash 0.0% $3.13 124 trace →
23 Claude Opus 4.6 0.0% $2.94 205 trace →
24 Claude Sonnet 4.6 0.0% $7.38 326 trace →

Click a row to replay how that model rebuilt this program, or the model name to open its full run