ComplexConstraints · Open Analysis

#	Model	Effort	Score	Cost / task	Harness
1	GPT-6 Astra	max	57.7%	–	–
2	Gemini 4	high	53.3%	–	–
3	Muse Spark 1.3	xhigh	51.9%	–	–
4	GPT-6 Sol	max	50.8%	–	–
5	GPT-5.6 Sol	max	50.5%	–	–
6	Opus 5.5	max	49.9%	–	–
7	DeepSeek V4.1 Flash	max	49.7%	–	–
8	GPT-5.5	xhigh	49.5%	–	–
9	Grok 4.7	xhigh	48.9%	–	–
10	Gemini 3.8 Flash	high	48.4%	–	–
11	GPT-5.4	xhigh	46.0%	–	–
12	Qwen 3.8 Max	xhigh	45.5%	–	–
13	Fable 5.1	max	45.1%	–	–
14	Gemini 3.1 Pro	high	43.7%	–	–
15	Qwen 3.8 Flash	xhigh	43.3%	–	–
16	Gemini 3.7 Flash	high	42.4%	–	–
17	DeepSeek V4 Pro	max	42.0%	–	–
18	Gemini 3.6 Flash	high	40.0%	–	–
19	DeepSeek V4 Flash	max	39.9%	–	–
20	Muse Spark 1.2	xhigh	39.9%	–	–
21	Fable 5	max	38.1%	–	–
22	Muse Spark 1.1	xhigh	38.1%	–	–
23	Kimi K3	max	37.9%	–	–
24	Opus 5	max	37.3%	–	–
25	Grok 4.6	xhigh	36.5%	–	–
26	Grok 4.5	high	35.9%	–	–
27	Opus 4.8	max	35.6%	–	–
28	Hy3	high	34.3%	–	–
29	Qwen 3.7 Max	max	33.5%	–	–
30	GLM 5.3	max	30.5%	–	–
31	Kimi K2.6	max	30.1%	–	–
32	GLM 5.2	max	29.3%	–	–
33	GLM 5.3 Flash	max	27.6%	–	–
34	Gemini 3.5 Flash-Lite	high	23.2%	–	–
35	Kimi K2.5	max	18.2%	–	–
36	Nemotron 3 Ultra 550B A55B	max	17.8%	–	–
37	Qwen 3.5 Plus	none	17.8%	–	–
38	Grok 4.20	xhigh	16.4%	–	–
39	Mistral Large 3	none	0.4%	–	–