Epistates/treemd | ProgramBench

A (TUI/CLI) markdown navigator with tree-based structural navigation.

1,569

Generated Behavioral Tests

85.1%

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)	85.1%	$51.59	294	trace →
2	Muse Spark 1.3 (max)	78.4%	$13.24	490	trace →
3	GPT-5.6 Sol (xhigh)	66.7%	$6.81	43	trace →
4	Muse Spark 1.3 (xhigh)	64.1%	$2.05	327	trace →
5	GPT 5.5 (xhigh)	58.9%	$7.32	69	trace →
6	Gemini 3.7 Flash	58.7%	$1.93	141	trace →
7	GLM-5.2	58.3%	$20.81	239	trace →
8	Claude Opus 4.8 (xhigh)	57.4%	$20.09	151	trace →
9	Muse Spark 1.2 (xhigh)	55.9%	$0.77	270	trace →
10	Claude Opus 4.6	55.1%	$20.28	431	trace →
11	Muse Spark 1.1 (xhigh)	54.2%	$0.38	236	trace →
12	Claude Sonnet 4.6	53.3%	$29.34	585	trace →
13	GPT 5.5 (high)	49.1%	$3.31	33	trace →
14	GPT-5.6 Sol	46.7%	$1.02	18	trace →
15	Claude Opus 4.7	45.4%	$2.54	66	trace →
16	GPT 5.5	45.3%	$1.62	20	trace →
17	Gemini 3.6 Flash	45.1%	$4.51	137	trace →
18	Gemini 3.5 Flash	45.1%	$6.72	243	trace →
19	Gemini 3.1 Pro	35.6%	$2.30	130	trace →
20	Claude Haiku 4.5	32.4%	$0.86	144	trace →
21	GPT 5.4	31.9%	$0.21	7	trace →
22	Gemini 3 Flash	27.5%	$0.31	117	trace →
23	GPT 5 mini	21.7%	$0.07	34	trace →
24	GPT 5.4 mini	14.6%	$0.04	14	trace →
25	Claude Opus 4.7 (xhigh)	1.5%	$5.66	147	trace →

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