pemistahl/grex | ProgramBench

A command-line tool and Rust library with Python bindings for generating regular expressions from user-provided test cases

1,312

Generated Behavioral Tests

99.9%

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)	99.9%	$24.33	243	trace →
2	GPT 6 Astra (max)	99.9%	$15.94	73	trace →
3	Muse Wonder (xhigh)	99.3%	$17.21	201	trace →
4	Claude Fable 5.1 (xhigh)	99.2%	$82.15	285	trace →
5	Muse Wonder Full	99.0%	TBA	349	trace →
6	Claude Opus 4.8 (xhigh)	98.1%	$17.62	125	trace →
7	GPT-5.6 Sol (xhigh)	97.4%	$13.41	78	trace →
8	GPT 5.5 (xhigh)	97.2%	$9.20	71	trace →
9	Muse Spark 1.3 (max)	94.1%	$5.89	296	trace →
10	Gemini 3.7 Flash	91.8%	$9.95	264	trace →
11	Muse Spark 1.1 (xhigh)	88.8%	$0.54	222	trace →
12	GPT 5.5 (high)	88.7%	$6.81	57	trace →
13	Gemini 3.5 Flash	84.4%	$6.38	205	trace →
14	Muse Spark 1.2 (xhigh)	84.2%	$8.73	486	trace →
15	Claude Opus 4.7 (xhigh)	82.1%	$31.59	231	trace →
16	Gemini 3.6 Flash	81.8%	$46.97	435	trace →
17	GPT-5.6 Sol	81.5%	$1.37	24	trace →
18	Claude Opus 4.7	73.9%	$7.62	153	trace →
19	GLM-5.2	70.9%	$40.95	292	trace →
20	GPT 5.4	66.5%	$0.32	7	trace →
21	Claude Opus 4.6	66.0%	$16.93	266	trace →
22	Gemini 3.1 Pro	61.4%	$1.29	159	trace →
23	GPT 5.5	58.1%	$1.56	22	trace →
24	Claude Sonnet 4.6	56.3%	$48.82	743	trace →
25	Gemini 3 Flash	52.8%	$0.23	81	trace →
26	GPT 5 mini	34.7%	$0.08	36	trace →
27	GPT 5.4 mini	21.6%	$0.05	10	trace →
28	Claude Opus 5 (xhigh)	0.0%	$17.74	116	trace →
29	Claude Haiku 4.5	0.0%	$1.02	104	trace →
30	Muse Spark 1.3 (xhigh)	n/a	$7.03	505	trace →

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