chmln/sd | ProgramBench

Intuitive find & replace CLI (sed alternative)

810

Generated Behavioral Tests

99.8%

Best Score

Muse Spark 1.3 (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)

25 runs

#	Model	Score	Cost	Calls
1	Muse Spark 1.3 (max)	99.8%	$2.77	163	trace →
2	Muse Spark 1.3 (xhigh)	99.5%	$0.49	109	trace →
3	Claude Opus 5 (xhigh)	99.4%	$19.17	174	trace →
4	GPT 5.5 (xhigh)	97.4%	$7.94	89	trace →
5	Claude Opus 4.8 (xhigh)	96.4%	$10.85	102	trace →
6	Claude Opus 4.7 (xhigh)	95.2%	$13.81	211	trace →
7	GPT-5.6 Sol (xhigh)	94.2%	$2.61	20	trace →
8	Gemini 3.7 Flash	93.5%	$1.02	131	trace →
9	Muse Spark 1.2 (xhigh)	92.7%	$0.39	122	trace →
10	Muse Spark 1.1 (xhigh)	92.5%	$0.09	125	trace →
11	GLM-5.2	92.2%	$14.17	190	trace →
12	GPT-5.6 Sol	91.0%	$0.84	17	trace →
13	Claude Opus 4.7	90.9%	$3.64	113	trace →
14	GPT 5.5 (high)	89.9%	$3.30	37	trace →
15	Claude Opus 4.6	89.9%	$7.07	270	trace →
16	Gemini 3.6 Flash	88.9%	$2.77	176	trace →
17	Gemini 3.5 Flash	88.8%	$1.50	118	trace →
18	GPT 5.5	86.0%	$0.87	17	trace →
19	Claude Sonnet 4.6	85.1%	$3.76	181	trace →
20	GPT 5.4	81.6%	$0.22	8	trace →
21	Gemini 3 Flash	80.7%	$0.25	111	trace →
22	Gemini 3.1 Pro	80.0%	$1.33	122	trace →
23	GPT 5 mini	70.9%	$0.03	22	trace →
24	Claude Haiku 4.5	68.6%	$0.87	149	trace →
25	GPT 5.4 mini	45.8%	$0.04	14	trace →

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