alexpovel/srgn | ProgramBench

A grep-like tool which understands source code syntax and allows for manipulation in addition to search

1,852

Generated Behavioral Tests

98.7%

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)	98.7%	$64.45	415	trace →
2	Muse Wonder Full	96.7%	TBA	310	trace →
3	GPT 6 Astra (max)	96.2%	$13.65	68	trace →
4	Claude Fable 5.1 (xhigh)	96.0%	$89.32	333	trace →
5	Muse Wonder (xhigh)	95.5%	$56.99	517	trace →
6	Claude Opus 5 (xhigh)	87.8%	$81.76	292	trace →
7	Muse Spark 1.3 (max)	85.5%	$3.70	254	trace →
8	GPT 5.5 (xhigh)	83.0%	$10.26	91	trace →
9	GPT 5.5 (high)	79.8%	$3.89	40	trace →
10	Claude Opus 4.8 (xhigh)	78.4%	$13.93	106	trace →
11	GPT-5.6 Sol (xhigh)	77.9%	$7.55	41	trace →
12	Muse Spark 1.2 (xhigh)	76.3%	$0.17	217	trace →
13	Muse Spark 1.1 (xhigh)	72.2%	$0.52	312	trace →
14	Gemini 3.6 Flash	71.7%	$3.25	139	trace →
15	Muse Spark 1.3 (xhigh)	70.8%	$2.17	268	trace →
16	Claude Sonnet 4.6	69.5%	$19.30	565	trace →
17	GPT 5.5	68.6%	$1.63	22	trace →
18	Gemini 3.5 Flash	67.7%	$6.42	237	trace →
19	GLM-5.2	67.2%	$32.54	308	trace →
20	GPT-5.6 Sol	67.0%	$1.33	16	trace →
21	Claude Opus 4.7 (xhigh)	65.9%	$6.38	118	trace →
22	Claude Opus 4.6	62.8%	$13.43	344	trace →
23	GPT 5.4	58.1%	$0.41	12	trace →
24	Gemini 3.1 Pro	57.9%	$1.13	94	trace →
25	Gemini 3 Flash	53.1%	$0.25	111	trace →
26	GPT 5 mini	46.0%	$0.04	13	trace →
27	Claude Haiku 4.5	43.6%	$1.15	167	trace →
28	Gemini 3.7 Flash	3.7%	$2.04	145	trace →
29	Claude Opus 4.7	2.4%	$1.57	64	trace →
30	GPT 5.4 mini	0.0%	$0.03	6	trace →

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