tree-sitter/tree-sitter

An incremental parsing system for programming tools

1,232

Generated Behavioral Tests

58.7%

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)	58.7%	$57.53	355	trace →
2	Muse Spark 1.3 (max)	47.6%	$3.11	236	trace →
3	Muse Spark 1.3 (xhigh)	47.2%	$1.21	211	trace →
4	Claude Opus 4.8 (xhigh)	46.5%	$17.26	145	trace →
5	GPT 5.5 (xhigh)	46.1%	$10.62	88	trace →
6	GPT-5.6 Sol (xhigh)	44.0%	$4.40	40	trace →
7	GPT 5.5 (high)	38.4%	$4.10	39	trace →
8	GLM-5.2	37.7%	$23.55	235	trace →
9	Claude Opus 4.7 (xhigh)	37.3%	$10.81	197	trace →
10	Claude Sonnet 4.6	37.2%	$15.00	402	trace →
11	Muse Spark 1.1 (xhigh)	36.9%	$0.25	93	trace →
12	Claude Opus 4.6	36.1%	$11.07	264	trace →
13	Muse Spark 1.2 (xhigh)	35.4%	$7.26	344	trace →
14	Claude Opus 4.7	35.1%	$8.38	139	trace →
15	GPT 5.5	34.2%	$1.04	19	trace →
16	GPT-5.6 Sol	30.6%	$1.00	15	trace →
17	GPT 5.4	29.1%	$0.28	9	trace →
18	Gemini 3.1 Pro	23.0%	$0.89	73	trace →
19	Claude Haiku 4.5	14.0%	$0.53	121	trace →
20	Gemini 3.6 Flash	13.4%	$2.60	144	trace →
21	Gemini 3.7 Flash	13.4%	$1.35	134	trace →
22	GPT 5 mini	13.4%	$0.03	19	trace →
23	Gemini 3.5 Flash	11.8%	$1.81	87	trace →
24	Gemini 3 Flash	6.9%	$0.30	90	trace →
25	GPT 5.4 mini	6.2%	$0.05	74	trace →

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