← Back to leaderboard  ·  Show all task instances

mgdm/htmlq

Like jq, but for HTML.

7,520 rs easy
1,455
Generated Behavioral Tests
99.4%
Best Score
Muse Spark 1.3 (max)

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 Muse Spark 1.3 (max) 99.4% $4.00 268 trace →
2 Muse Spark 1.3 (xhigh) 98.1% $1.04 209 trace →
3 GPT 5.5 (xhigh) 98.0% $5.09 69 trace →
4 GPT-5.6 Sol (xhigh) 97.4% $2.54 22 trace →
5 Claude Opus 4.8 (xhigh) 96.6% $16.43 115 trace →
6 GPT 5.5 (high) 96.4% $3.62 52 trace →
7 Gemini 3.7 Flash 96.4% $1.48 134 trace →
8 Claude Opus 4.7 (xhigh) 96.2% $15.99 192 trace →
9 Gemini 3.5 Flash 95.4% $5.29 149 trace →
10 GPT-5.6 Sol 95.1% $0.65 14 trace →
11 Muse Spark 1.2 (xhigh) 94.0% $1.48 197 trace →
12 Claude Opus 4.6 93.9% $13.55 325 trace →
13 GLM-5.2 93.8% $20.51 170 trace →
14 Gemini 3.6 Flash 92.3% $4.14 213 trace →
15 Claude Opus 4.7 92.3% $3.11 108 trace →
16 Claude Sonnet 4.6 88.0% $22.96 627 trace →
17 Gemini 3.1 Pro 87.6% $1.46 104 trace →
18 Gemini 3 Flash 84.4% $0.29 140 trace →
19 GPT 5.5 84.3% $0.99 17 trace →
20 Claude Haiku 4.5 78.0% $0.58 144 trace →
21 GPT 5.4 60.8% $0.30 14 trace →
22 GPT 5.4 mini 37.8% $0.03 9 trace →
23 GPT 5 mini 1.2% $0.02 13 trace →
24 Claude Opus 5 (xhigh) 0.0% $4.32 67 trace →

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