← Back to leaderboard  ·  Show all task instances

typst/typst

A markup-based typesetting system that is powerful and easy to learn.

52,957 rs hard
1,724
Generated Behavioral Tests
52.8%
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) 52.8% $23.63 190 trace →
2 Muse Spark 1.3 (xhigh) 47.5% $0.88 177 trace →
3 GPT-5.6 Sol (xhigh) 44.7% $5.12 29 trace →
4 GLM-5.2 42.0% $34.69 276 trace →
5 Muse Spark 1.2 (xhigh) 36.9% $1.53 123 trace →
6 GPT-5.6 Sol 34.9% $0.66 10 trace →
7 Gemini 3.6 Flash 32.7% $2.25 104 trace →
8 GPT 5.5 32.7% $0.75 13 trace →
9 Muse Spark 1.3 (max) 28.7% $15.60 651 trace →
10 GPT 5.4 28.0% $0.39 11 trace →
11 Claude Haiku 4.5 20.6% $0.72 127 trace →
12 Claude Opus 4.8 (xhigh) 16.8% $20.14 159 trace →
13 Gemini 3 Flash 15.8% $0.28 75 trace →
14 Gemini 3.1 Pro 15.5% $0.67 44 trace →
15 Claude Opus 4.7 (xhigh) 14.4% $5.44 125 trace →
16 Claude Opus 4.6 13.3% $4.62 285 trace →
17 Claude Opus 4.7 11.7% $1.51 50 trace →
18 GPT 5 mini 3.8% $0.02 8 trace →
19 GPT 5.5 (xhigh) 3.7% $10.06 83 trace →
20 Gemini 3.7 Flash 3.7% $1.12 126 trace →
21 Muse Spark 1.1 (xhigh) 3.7% $0.91 172 trace →
22 Gemini 3.5 Flash 3.2% $2.50 93 trace →
23 GPT 5.4 mini 2.7% $0.03 7 trace →
24 GPT 5.5 (high) 1.3% $3.71 32 trace →
25 Claude Sonnet 4.6 0.0% $52.50 855 trace →

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