tarka/xcp | ProgramBench

An extended `cp`

1,184

Generated Behavioral Tests

99.6%

Best Score

GPT 6 Astra (max)

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)

28 runs

#	Model	Score	Cost	Calls
1	GPT 6 Astra (max)	99.6%	$7.55	53	trace →
2	Claude Opus 5 (xhigh)	99.5%	$50.85	289	trace →
3	Claude Opus 5.5 (xhigh)	99.3%	$17.34	259	trace →
4	Muse Wonder Full	99.3%	TBA	234	trace →
5	Claude Fable 5.1 (xhigh)	99.2%	$51.94	229	trace →
6	Muse Wonder (xhigh)	99.2%	$17.16	208	trace →
7	Muse Spark 1.3 (max)	99.2%	$4.49	274	trace →
8	GPT-5.6 Sol (xhigh)	98.1%	$4.77	34	trace →
9	Claude Opus 4.8 (xhigh)	96.9%	$8.45	87	trace →
10	Gemini 3.7 Flash	95.9%	$1.22	122	trace →
11	GPT 5.5 (xhigh)	95.8%	$6.37	62	trace →
12	Claude Opus 4.6	92.6%	$5.01	138	trace →
13	Muse Spark 1.1 (xhigh)	92.4%	$0.36	107	trace →
14	Claude Sonnet 4.6	91.7%	$8.99	330	trace →
15	Claude Opus 4.7 (xhigh)	91.4%	$13.71	263	trace →
16	GPT 5.4	84.5%	$0.28	10	trace →
17	Gemini 3.1 Pro	73.7%	$0.72	42	trace →
18	GLM-5.2	39.5%	$9.98	171	trace →
19	Muse Spark 1.2 (xhigh)	39.5%	$2.26	252	trace →
20	GPT 5.5	38.7%	$1.33	18	trace →
21	Gemini 3.6 Flash	38.0%	$2.84	136	trace →
22	Claude Opus 4.7	37.2%	$2.57	85	trace →
23	GPT 5.4 mini	22.5%	$0.04	18	trace →
24	Gemini 3 Flash	9.0%	$0.24	57	trace →
25	GPT 5 mini	4.3%	$0.02	16	trace →
26	Muse Spark 1.3 (xhigh)	0.0%	$1.10	199	trace →
27	Claude Haiku 4.5	0.0%	$0.73	107	trace →
28	GPT 5.5 (high)	n/a	$1.95	25	trace →

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