← Back to leaderboard  ·  Show all task instances

anordal/shellharden

The corrective bash syntax highlighter

4,778 rs easy
1,095
Generated Behavioral Tests
96.7%
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

25 runs
# Model Score Cost Calls
1 Claude Opus 4.8 (xhigh) 96.7% $63.32 271 trace →
2 GPT-5.6 Sol (xhigh) 91.8% $6.50 50 trace →
3 Muse Spark 1.3 (xhigh) 90.1% $2.21 228 trace →
4 Gemini 3.7 Flash 89.2% $12.59 257 trace →
5 Muse Spark 1.3 (max) 88.6% $6.67 355 trace →
6 GPT 5.5 (xhigh) 88.2% $11.50 93 trace →
7 GPT 5.5 (high) 87.1% $4.26 42 trace →
8 Claude Opus 4.7 (xhigh) 82.8% $11.64 179 trace →
9 GPT-5.6 Sol 81.2% $0.47 13 trace →
10 Gemini 3.6 Flash 80.2% $11.64 216 trace →
11 GPT 5.5 79.7% $1.16 18 trace →
12 Muse Spark 1.1 (xhigh) 79.2% $0.53 277 trace →
13 Gemini 3.5 Flash 78.4% $3.14 117 trace →
14 GLM-5.2 75.1% $37.58 280 trace →
15 Claude Opus 4.7 72.0% $1.99 69 trace →
16 Gemini 3.1 Pro 71.1% $1.40 143 trace →
17 GPT 5.4 63.3% $0.16 9 trace →
18 Gemini 3 Flash 59.6% $0.24 90 trace →
19 Claude Haiku 4.5 54.4% $0.96 147 trace →
20 Claude Opus 5 (xhigh) 51.3% $112.85 403 trace →
21 Claude Sonnet 4.6 42.9% $74.50 840 trace →
22 Muse Spark 1.2 (xhigh) 41.8% $1.44 464 trace →
23 GPT 5.4 mini 16.2% $0.04 18 trace →
24 GPT 5 mini 7.2% $0.03 14 trace →
25 Claude Opus 4.6 0.0% $2.13 256 trace →

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