svenstaro/genact | ProgramBench

🌀 A nonsense activity generator

231

Generated Behavioral Tests

99.1%

Best Score

Muse Wonder Full

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	Muse Wonder Full	99.1%	TBA	388	trace →
2	Claude Opus 5.5 (xhigh)	98.7%	$41.45	398	trace →
3	Muse Wonder (xhigh)	98.7%	$34.53	389	trace →
4	Claude Fable 5.1 (xhigh)	95.7%	$96.23	392	trace →
5	GPT 6 Astra (max)	88.3%	$7.02	46	trace →
6	Claude Opus 5 (xhigh)	86.6%	$67.92	355	trace →
7	Muse Spark 1.3 (max)	82.3%	$3.48	251	trace →
8	GPT-5.6 Sol (xhigh)	74.9%	$3.29	26	trace →
9	GLM-5.2	74.9%	$56.55	423	trace →
10	Claude Opus 4.8 (xhigh)	73.2%	$23.01	185	trace →
11	GPT 5.5 (xhigh)	70.1%	$5.22	42	trace →
12	Gemini 3.5 Flash	68.8%	$5.23	115	trace →
13	Gemini 3.7 Flash	61.9%	$2.33	113	trace →
14	Claude Sonnet 4.6	58.4%	$17.93	298	trace →
15	Muse Spark 1.1 (xhigh)	57.1%	$1.20	181	trace →
16	Muse Spark 1.3 (xhigh)	54.5%	$0.97	177	trace →
17	Muse Spark 1.2 (xhigh)	54.5%	$0.66	100	trace →
18	Claude Opus 4.6	52.4%	$11.11	158	trace →
19	Gemini 3.6 Flash	49.8%	$2.87	93	trace →
20	Claude Opus 4.7 (xhigh)	48.9%	$3.83	71	trace →
21	GPT 5.5 (high)	35.9%	$2.91	24	trace →
22	Gemini 3.1 Pro	35.5%	$1.25	44	trace →
23	GPT 5.4	24.2%	$0.46	10	trace →
24	GPT 5.5	23.4%	$0.94	13	trace →
25	Claude Haiku 4.5	22.9%	$0.58	79	trace →
26	Claude Opus 4.7	22.1%	$1.70	49	trace →
27	Gemini 3 Flash	20.8%	$0.47	48	trace →
28	GPT-5.6 Sol	17.7%	$0.37	8	trace →
29	GPT 5.4 mini	12.6%	$0.02	10	trace →
30	GPT 5 mini	1.7%	$0.02	9	trace →

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