hush-shell/hush | ProgramBench

Hush is a unix shell based on the Lua programming language

1,201

Generated Behavioral Tests

98.6%

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)	98.6%	$47.05	429	trace →
2	Claude Fable 5.1 (xhigh)	98.3%	$65.26	341	trace →
3	Muse Wonder Full	97.2%	TBA	317	trace →
4	Muse Wonder (xhigh)	96.8%	$46.50	527	trace →
5	Claude Opus 5 (xhigh)	96.6%	$71.83	422	trace →
6	GPT 6 Astra (max)	96.4%	$9.61	55	trace →
7	GPT-5.6 Sol (xhigh)	90.3%	$11.19	68	trace →
8	Claude Opus 4.8 (xhigh)	89.3%	$51.55	301	trace →
9	Gemini 3.7 Flash	87.9%	$4.60	178	trace →
10	Muse Spark 1.3 (max)	87.2%	$9.53	401	trace →
11	Claude Opus 4.6	82.3%	$16.84	374	trace →
12	Claude Sonnet 4.6	75.8%	$67.34	1,000	trace →
13	GLM-5.2	74.5%	$37.89	375	trace →
14	GPT 5.5 (high)	67.4%	$5.33	56	trace →
15	Gemini 3.6 Flash	61.8%	$14.23	284	trace →
16	Claude Opus 4.7 (xhigh)	36.4%	$2.15	57	trace →
17	Gemini 3.5 Flash	35.1%	$53.40	626	trace →
18	Muse Spark 1.3 (xhigh)	26.8%	$8.19	588	trace →
19	Muse Spark 1.1 (xhigh)	24.5%	$2.09	238	trace →
20	Claude Opus 4.7	15.0%	$0.75	28	trace →
21	Gemini 3.1 Pro	14.7%	$1.01	61	trace →
22	GPT 5.5	12.3%	$1.83	33	trace →
23	GPT-5.6 Sol	12.2%	$1.32	22	trace →
24	Gemini 3 Flash	8.4%	$0.48	158	trace →
25	GPT 5.4 mini	7.9%	$0.03	19	trace →
26	GPT 5 mini	6.7%	$0.03	12	trace →
27	GPT 5.4	5.8%	$0.23	10	trace →
28	Claude Haiku 4.5	4.4%	$0.72	87	trace →
29	GPT 5.5 (xhigh)	0.8%	$15.30	119	trace →
30	Muse Spark 1.2 (xhigh)	0.0%	$3.92	227	trace →

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