ast-grep/ast-grep | ProgramBench

⚡A CLI tool for code structural search, lint and rewriting. Written in Rust

877

Generated Behavioral Tests

77.2%

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)	77.2%	$52.25	319	trace →
2	Claude Fable 5.1 (xhigh)	71.4%	$57.15	274	trace →
3	Muse Wonder Full	68.0%	TBA	539	trace →
4	GPT 6 Astra (max)	67.4%	$13.55	75	trace →
5	Claude Opus 5 (xhigh)	57.2%	$42.34	228	trace →
6	Muse Spark 1.3 (max)	45.2%	$3.96	358	trace →
7	GPT-5.6 Sol (xhigh)	32.7%	$5.19	39	trace →
8	Muse Spark 1.3 (xhigh)	28.2%	$1.67	293	trace →
9	Gemini 3.7 Flash	27.6%	$1.38	141	trace →
10	Gemini 3.6 Flash	24.5%	$3.30	143	trace →
11	GLM-5.2	23.9%	$24.06	210	trace →
12	Muse Spark 1.1 (xhigh)	23.8%	$0.14	157	trace →
13	GPT 5.5 (xhigh)	23.7%	$8.63	67	trace →
14	GPT 5.5 (high)	20.3%	$3.93	35	trace →
15	Muse Spark 1.2 (xhigh)	18.7%	$0.12	143	trace →
16	Claude Opus 4.8 (xhigh)	17.3%	$13.83	116	trace →
17	GPT 5.5	12.3%	$1.48	21	trace →
18	Claude Opus 4.6	11.7%	$11.94	322	trace →
19	GPT-5.6 Sol	10.6%	$0.73	11	trace →
20	Claude Opus 4.7 (xhigh)	9.6%	$5.83	123	trace →
21	GPT 5.4	9.2%	$0.41	11	trace →
22	Claude Sonnet 4.6	8.1%	$28.64	621	trace →
23	Gemini 3.5 Flash	7.8%	$6.16	138	trace →
24	Gemini 3.1 Pro	5.8%	$0.79	66	trace →
25	Claude Opus 4.7	3.9%	$1.45	46	trace →
26	Gemini 3 Flash	2.6%	$0.27	70	trace →
27	Claude Haiku 4.5	2.6%	$0.49	92	trace →
28	GPT 5 mini	1.4%	$0.04	16	trace →
29	Muse Wonder (xhigh)	0.0%	$61.76	549	trace →
30	GPT 5.4 mini	0.0%	$0.05	38	trace →

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