pls-rs/pls | ProgramBench

pls is a prettier and powerful ls(1) for the pros.

331

Generated Behavioral Tests

94.9%

Best Score

Claude Opus 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)

25 runs

#	Model	Score	Cost	Calls
1	Claude Opus 5 (xhigh)	94.9%	$53.59	297	trace →
2	Claude Opus 4.8 (xhigh)	86.7%	$48.46	261	trace →
3	Muse Spark 1.3 (max)	78.9%	$5.27	349	trace →
4	GPT 5.5 (xhigh)	76.4%	$10.30	119	trace →
5	GPT-5.6 Sol (xhigh)	75.8%	$8.64	63	trace →
6	Muse Spark 1.3 (xhigh)	72.5%	$1.83	246	trace →
7	Gemini 3.7 Flash	72.5%	$1.75	120	trace →
8	GLM-5.2	69.5%	$32.84	277	trace →
9	GPT 5.5 (high)	67.4%	$3.10	43	trace →
10	Muse Spark 1.2 (xhigh)	64.7%	$3.85	286	trace →
11	Gemini 3.1 Pro	62.2%	$1.60	127	trace →
12	GPT-5.6 Sol	58.9%	$1.24	17	trace →
13	Claude Sonnet 4.6	58.6%	$40.54	680	trace →
14	Gemini 3.6 Flash	51.4%	$3.58	143	trace →
15	Claude Opus 4.7 (xhigh)	47.4%	$7.41	155	trace →
16	Muse Spark 1.1 (xhigh)	42.6%	$0.16	97	trace →
17	Gemini 3.5 Flash	40.8%	$5.98	150	trace →
18	Claude Opus 4.6	30.8%	$23.17	497	trace →
19	GPT 5.5	29.9%	$0.89	11	trace →
20	Claude Opus 4.7	24.2%	$1.86	75	trace →
21	GPT 5.4	16.9%	$0.24	9	trace →
22	Claude Haiku 4.5	16.3%	$0.77	112	trace →
23	Gemini 3 Flash	13.9%	$0.35	78	trace →
24	GPT 5.4 mini	9.7%	$0.03	9	trace →
25	GPT 5 mini	2.4%	$0.01	7	trace →

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