astro/deadnix | ProgramBench

Scan Nix files for dead code

602

Generated Behavioral Tests

99.5%

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)

30 runs

#	Model	Score	Cost	Calls
1	Claude Opus 5 (xhigh)	99.5%	$101.73	400	trace →
2	Claude Opus 5.5 (xhigh)	99.3%	$22.11	192	trace →
3	Claude Fable 5.1 (xhigh)	99.3%	$48.99	243	trace →
4	Muse Wonder (xhigh)	99.3%	$29.51	301	trace →
5	GPT 6 Astra (max)	99.2%	$10.09	46	trace →
6	Muse Spark 1.3 (max)	98.7%	$4.81	288	trace →
7	Muse Wonder Full	98.5%	TBA	206	trace →
8	GPT-5.6 Sol (xhigh)	96.8%	$8.15	49	trace →
9	Claude Opus 4.8 (xhigh)	94.2%	$23.32	118	trace →
10	Gemini 3.7 Flash	92.4%	$2.19	131	trace →
11	Muse Spark 1.3 (xhigh)	92.0%	$1.38	182	trace →
12	GPT 5.5 (xhigh)	91.4%	$9.41	58	trace →
13	GLM-5.2	91.4%	$17.01	191	trace →
14	Claude Opus 4.7 (xhigh)	89.7%	$19.53	167	trace →
15	Gemini 3.6 Flash	86.9%	$5.78	154	trace →
16	Claude Opus 4.7	85.5%	$9.59	142	trace →
17	GPT-5.6 Sol	82.9%	$1.46	15	trace →
18	Claude Sonnet 4.6	81.1%	$12.78	294	trace →
19	GPT 5.5	79.7%	$1.74	20	trace →
20	Gemini 3.5 Flash	78.7%	$4.35	98	trace →
21	Muse Spark 1.1 (xhigh)	71.8%	$0.50	316	trace →
22	Gemini 3.1 Pro	70.1%	$1.71	81	trace →
23	Gemini 3 Flash	57.3%	$0.27	57	trace →
24	Claude Haiku 4.5	44.9%	$1.23	122	trace →
25	GPT 5.4	44.5%	$0.40	9	trace →
26	GPT 5 mini	23.1%	$0.05	8	trace →
27	GPT 5.5 (high)	6.5%	$4.79	43	trace →
28	GPT 5.4 mini	2.0%	$0.05	29	trace →
29	Muse Spark 1.2 (xhigh)	1.5%	$1.66	403	trace →
30	Claude Opus 4.6	0.0%	$3.29	258	trace →

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