typst/typst | ProgramBench

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

1,724

Generated Behavioral Tests

79.9%

Best Score

Claude Opus 5.5 (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)

30 runs

#	Model	Score	Cost	Calls
1	Claude Opus 5.5 (xhigh)	79.9%	$73.14	370	trace →
2	Claude Fable 5.1 (xhigh)	77.3%	$113.75	295	trace →
3	GPT 6 Astra (max)	69.1%	$12.48	68	trace →
4	Muse Wonder (xhigh)	69.0%	$40.90	391	trace →
5	Muse Wonder Full	61.0%	TBA	160	trace →
6	Claude Opus 5 (xhigh)	52.8%	$23.63	190	trace →
7	Muse Spark 1.3 (xhigh)	47.5%	$0.88	177	trace →
8	GPT-5.6 Sol (xhigh)	44.7%	$5.12	29	trace →
9	GLM-5.2	42.0%	$34.69	276	trace →
10	Muse Spark 1.2 (xhigh)	36.9%	$1.53	123	trace →
11	GPT-5.6 Sol	34.9%	$0.66	10	trace →
12	Gemini 3.6 Flash	32.7%	$2.25	104	trace →
13	GPT 5.5	32.7%	$0.75	13	trace →
14	Muse Spark 1.3 (max)	28.7%	$15.60	651	trace →
15	GPT 5.4	28.0%	$0.39	11	trace →
16	Claude Haiku 4.5	20.6%	$0.72	127	trace →
17	Claude Opus 4.8 (xhigh)	16.8%	$20.14	159	trace →
18	Gemini 3 Flash	15.8%	$0.28	75	trace →
19	Gemini 3.1 Pro	15.5%	$0.67	44	trace →
20	Claude Opus 4.7 (xhigh)	14.4%	$5.44	125	trace →
21	Claude Opus 4.6	13.3%	$4.62	285	trace →
22	Claude Opus 4.7	11.7%	$1.51	50	trace →
23	GPT 5 mini	3.8%	$0.02	8	trace →
24	GPT 5.5 (xhigh)	3.7%	$10.06	83	trace →
25	Gemini 3.7 Flash	3.7%	$1.12	126	trace →
26	Muse Spark 1.1 (xhigh)	3.7%	$0.91	172	trace →
27	Gemini 3.5 Flash	3.2%	$2.50	93	trace →
28	GPT 5.4 mini	2.7%	$0.03	7	trace →
29	GPT 5.5 (high)	1.3%	$3.71	32	trace →
30	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