Lymphatus/caesium-clt | ProgramBench

Caesium Command Line Tools - Lossy/lossless image compression tool

575

Generated Behavioral Tests

99.8%

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)	99.8%	$50.69	365	trace →
2	GPT 6 Astra (max)	99.5%	$23.62	75	trace →
3	Claude Fable 5.1 (xhigh)	99.0%	$44.68	275	trace →
4	Muse Wonder Full	99.0%	TBA	446	trace →
5	Muse Wonder (xhigh)	99.0%	$28.06	321	trace →
6	Claude Opus 5 (xhigh)	96.5%	$32.05	212	trace →
7	GPT-5.6 Sol (xhigh)	96.2%	$3.88	33	trace →
8	GPT 5.5 (xhigh)	96.2%	$7.95	71	trace →
9	Gemini 3.7 Flash	95.8%	$2.29	128	trace →
10	Muse Spark 1.3 (xhigh)	95.7%	$0.87	174	trace →
11	GPT 5.5 (high)	92.5%	$5.06	60	trace →
12	Claude Opus 4.6	92.3%	$11.47	294	trace →
13	GLM-5.2	85.0%	$16.72	163	trace →
14	Muse Spark 1.2 (xhigh)	84.9%	$1.60	148	trace →
15	Gemini 3.5 Flash	83.5%	$5.51	174	trace →
16	Claude Opus 4.8 (xhigh)	80.9%	$9.88	85	trace →
17	Muse Spark 1.1 (xhigh)	80.2%	$1.69	220	trace →
18	Claude Opus 4.7 (xhigh)	76.5%	$9.10	164	trace →
19	GPT 5.5	76.3%	$1.35	23	trace →
20	Claude Opus 4.7	73.6%	$1.89	57	trace →
21	GPT-5.6 Sol	72.9%	$1.07	17	trace →
22	Gemini 3.1 Pro	57.9%	$1.54	95	trace →
23	Claude Haiku 4.5	57.6%	$0.86	135	trace →
24	Gemini 3 Flash	49.4%	$0.40	139	trace →
25	Muse Spark 1.3 (max)	40.7%	$5.03	298	trace →
26	GPT 5.4 mini	17.2%	$0.04	28	trace →
27	GPT 5 mini	12.5%	$0.02	12	trace →
28	Claude Sonnet 4.6	5.7%	$8.73	350	trace →
29	Gemini 3.6 Flash	1.6%	$6.01	203	trace →
30	GPT 5.4	1.6%	$0.61	17	trace →

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