← Back to leaderboard  ·  Show all task instances

dandavison/delta

A syntax-highlighting pager for git, diff, grep, rg --json, and blame output

30,445 rs hard
949
Generated Behavioral Tests
66.5%
Best Score
Claude Opus 5 (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 5 (xhigh) 66.5% $92.02 371 trace →
2 GPT-5.6 Sol (xhigh) 47.7% $10.19 41 trace →
3 GPT 5.5 (xhigh) 39.0% $10.75 65 trace →
4 Claude Sonnet 4.6 37.3% $34.08 470 trace →
5 Muse Spark 1.3 (xhigh) 36.9% $1.43 167 trace →
6 GLM-5.2 36.8% $31.00 275 trace →
7 Claude Opus 4.8 (xhigh) 35.8% $14.66 131 trace →
8 Muse Spark 1.2 (xhigh) 34.1% $0.68 142 trace →
9 GPT 5.5 (high) 31.8% $6.94 57 trace →
10 GPT 5.5 30.1% $0.96 14 trace →
11 GPT-5.6 Sol 29.4% $0.93 10 trace →
12 Claude Opus 4.7 (xhigh) 28.1% $1.19 48 trace →
13 Claude Opus 4.7 23.1% $0.35 22 trace →
14 Gemini 3 Flash 21.3% $0.20 52 trace →
15 GPT 5 mini 20.8% $0.03 14 trace →
16 GPT 5.4 17.8% $0.23 7 trace →
17 Claude Haiku 4.5 10.1% $1.28 166 trace →
18 Gemini 3.6 Flash 9.6% $3.73 129 trace →
19 GPT 5.4 mini 5.4% $0.05 8 trace →
20 Muse Spark 1.1 (xhigh) 0.8% $0.39 186 trace →
21 Claude Opus 4.6 0.8% $8.32 236 trace →
22 Muse Spark 1.3 (max) 0.7% $12.41 370 trace →
23 Gemini 3.7 Flash 0.7% $3.80 141 trace →
24 Gemini 3.5 Flash 0.7% $1.85 79 trace →
25 Gemini 3.1 Pro 0.0% $1.70 120 trace →

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