← Back to leaderboard  ·  Show all task instances

OSGeo/gdal

GDAL is an open source MIT licensed translator library for raster and vector geospatial data formats.

5,875 cpp
657
Generated Behavioral Tests
39.0%
Best Score
GPT-5.6 Sol (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 GPT-5.6 Sol (xhigh) 39.0% $6.67 39 trace →
2 Muse Spark 1.1 (xhigh) 21.6% $1.25 227 trace →
3 Claude Opus 5 (xhigh) 19.2% $39.61 252 trace →
4 GPT 5.5 (high) 17.2% $3.46 36 trace →
5 Gemini 3.6 Flash 16.0% $7.02 181 trace →
6 Claude Opus 4.8 (xhigh) 12.8% $17.95 146 trace →
7 GPT 5.5 11.9% $0.94 14 trace →
8 Claude Opus 4.7 (xhigh) 10.0% $8.27 154 trace →
9 Muse Spark 1.3 (xhigh) 9.3% $0.81 149 trace →
10 Claude Sonnet 4.6 7.5% $48.10 684 trace →
11 Gemini 3 Flash 6.7% $0.24 70 trace →
12 Claude Haiku 4.5 6.1% $0.64 142 trace →
13 GPT 5.4 5.3% $0.36 9 trace →
14 Gemini 3.1 Pro 5.0% $0.74 39 trace →
15 GPT 5.4 mini 5.0% $0.03 7 trace →
16 Muse Spark 1.3 (max) 4.0% $14.02 504 trace →
17 GLM-5.2 3.8% $6.12 97 trace →
18 GPT-5.6 Sol 3.7% $1.32 17 trace →
19 Gemini 3.7 Flash 3.7% $1.17 122 trace →
20 Claude Opus 4.7 3.7% $0.47 22 trace →
21 Muse Spark 1.2 (xhigh) 3.3% $4.73 216 trace →
22 GPT 5 mini 1.7% $0.09 35 trace →
23 GPT 5.5 (xhigh) 0.0% $4.39 41 trace →
24 Gemini 3.5 Flash 0.0% $4.95 163 trace →
25 Claude Opus 4.6 0.0% $2.60 218 trace →

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