bensadeh/tailspin | ProgramBench

🌀 A log file highlighter

614

Generated Behavioral Tests

95.3%

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)	95.3%	$52.38	298	trace →
2	Muse Spark 1.3 (max)	93.2%	$5.92	323	trace →
3	GPT-5.6 Sol (xhigh)	91.5%	$9.04	48	trace →
4	GPT 5.5 (xhigh)	88.1%	$15.26	97	trace →
5	Gemini 3.7 Flash	86.8%	$4.00	188	trace →
6	Muse Spark 1.2 (xhigh)	84.9%	$0.29	186	trace →
7	Claude Opus 4.8 (xhigh)	84.5%	$24.91	150	trace →
8	Gemini 3.6 Flash	82.2%	$10.86	221	trace →
9	GPT 5.5 (high)	78.0%	$3.64	33	trace →
10	Claude Sonnet 4.6	75.7%	$58.91	840	trace →
11	Claude Opus 4.7	75.6%	$2.18	65	trace →
12	GLM-5.2	74.7%	$49.10	385	trace →
13	Claude Opus 4.7 (xhigh)	73.9%	$26.95	265	trace →
14	Claude Opus 4.6	72.0%	$20.30	324	trace →
15	Gemini 3.5 Flash	70.8%	$5.93	157	trace →
16	GPT 5.5	68.4%	$1.45	24	trace →
17	GPT 5.4 mini	62.9%	$0.10	10	trace →
18	GPT 5.4	59.6%	$0.26	8	trace →
19	Gemini 3 Flash	56.8%	$0.67	130	trace →
20	GPT-5.6 Sol	32.4%	$1.07	17	trace →
21	GPT 5 mini	26.2%	$0.02	13	trace →
22	Muse Spark 1.1 (xhigh)	2.6%	$0.13	100	trace →
23	Gemini 3.1 Pro	2.1%	$1.56	51	trace →
24	Muse Spark 1.3 (xhigh)	0.0%	$2.80	262	trace →
25	Claude Haiku 4.5	0.0%	$0.87	101	trace →

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