duckdb/duckdb | ProgramBench

DuckDB is an analytical in-process SQL database management system

5,649

Generated Behavioral Tests

53.1%

Best Score

Claude Opus 5.5 (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 5.5 (xhigh)	53.1%	$1.51	59	trace →
2	Claude Fable 5.1 (xhigh)	52.0%	$102.08	316	trace →
3	GPT 6 Astra (max)	45.4%	$14.29	60	trace →
4	Muse Wonder Full	40.5%	TBA	501	trace →
5	Muse Wonder (xhigh)	38.1%	$47.76	395	trace →
6	Claude Opus 5 (xhigh)	27.1%	$45.12	243	trace →
7	GPT-5.6 Sol (xhigh)	22.9%	$11.19	83	trace →
8	Muse Spark 1.3 (max)	21.7%	$3.57	207	trace →
9	Muse Spark 1.3 (xhigh)	17.8%	$0.98	170	trace →
10	GPT 5.5 (xhigh)	16.8%	$7.51	77	trace →
11	GLM-5.2	14.8%	$21.10	188	trace →
12	Muse Spark 1.2 (xhigh)	12.9%	$1.12	227	trace →
13	GPT 5.5 (high)	12.8%	$3.92	35	trace →
14	Claude Opus 4.6	12.4%	$17.56	441	trace →
15	GPT-5.6 Sol	12.3%	$0.85	14	trace →
16	GPT 5.5	11.9%	$0.99	15	trace →
17	GPT 5.4	10.9%	$0.43	13	trace →
18	Gemini 3.5 Flash	9.4%	$5.55	188	trace →
19	Gemini 3.1 Pro	6.2%	$1.01	88	trace →
20	Gemini 3 Flash	4.7%	$0.49	118	trace →
21	GPT 5.4 mini	4.3%	$0.03	57	trace →
22	Claude Opus 4.7 (xhigh)	4.1%	$0.70	32	trace →
23	Claude Opus 4.7	4.0%	$1.41	62	trace →
24	GPT 5 mini	4.0%	$0.03	14	trace →
25	Gemini 3.6 Flash	3.8%	$2.89	138	trace →
26	Gemini 3.7 Flash	3.8%	$0.88	106	trace →
27	Claude Haiku 4.5	3.8%	$0.86	121	trace →
28	Claude Opus 4.8 (xhigh)	3.3%	$28.66	170	trace →
29	Muse Spark 1.1 (xhigh)	3.3%	$0.52	207	trace →
30	Claude Sonnet 4.6	0.0%	$8.24	288	trace →

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