konradsz/igrep | ProgramBench

Interactive Grep

385

Generated Behavioral Tests

92.9%

Best Score

Claude Opus 4.8 (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 4.8 (xhigh)	92.9%	$32.24	202	trace →
2	Muse Spark 1.3 (xhigh)	92.5%	$2.52	333	trace →
3	Claude Opus 5 (xhigh)	87.8%	$30.15	204	trace →
4	GPT 5.5 (xhigh)	86.5%	$4.92	51	trace →
5	Muse Spark 1.3 (max)	85.7%	$4.45	339	trace →
6	GLM-5.2	84.9%	$22.32	200	trace →
7	Muse Spark 1.2 (xhigh)	79.2%	$0.18	106	trace →
8	Gemini 3.6 Flash	76.6%	$2.80	105	trace →
9	GPT 5.5 (high)	76.1%	$2.19	32	trace →
10	Claude Opus 4.7 (xhigh)	75.6%	$8.64	125	trace →
11	Gemini 3.7 Flash	75.1%	$1.33	85	trace →
12	GPT 5.5	74.0%	$1.02	20	trace →
13	Claude Opus 4.7	73.5%	$1.38	54	trace →
14	Gemini 3.5 Flash	72.2%	$3.56	107	trace →
15	Gemini 3.1 Pro	61.8%	$0.74	53	trace →
16	Claude Sonnet 4.6	60.5%	$10.26	279	trace →
17	Gemini 3 Flash	56.9%	$0.21	49	trace →
18	Claude Haiku 4.5	53.5%	$0.43	76	trace →
19	GPT-5.6 Sol (xhigh)	51.9%	$3.24	18	trace →
20	GPT 5 mini	51.7%	$0.04	12	trace →
21	GPT-5.6 Sol	48.8%	$0.89	13	trace →
22	GPT 5.4 mini	39.2%	$0.03	12	trace →
23	GPT 5.4	8.8%	$0.21	14	trace →
24	Muse Spark 1.1 (xhigh)	0.0%	$0.76	168	trace →
25	Claude Opus 4.6	0.0%	$1.32	185	trace →

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