Judgemark v4 · Open Analysis

#	Model	Effort	Score	Cost / task	Harness
1	Opus 4.6	max	91	–	–
2	GPT-5.5	xhigh	88	–	–
3	Opus 4.7	max	84	–	–
4	Gemini 3.7 Flash	max	83	–	–
5	Sonnet 4.6	max	82	–	–
6	Muse Spark 1.2	max	79	–	–
7	Opus 5	max	79	–	–
8	Gemini 3.1 Pro	high	79	–	–
9	Opus 4.8	max	78	–	–
10	Grok 4.5	max	77	–	–
11	GLM 5.2	max	73	–	–
12	GLM 5.3	max	73	–	–
13	Gemma 4 31B IT	high	72	–	–
14	GPT-5.4	xhigh	72	–	–
15	Sonnet 5	max	71	–	–
16	Gemini 3.5 Flash	xhigh	69	–	–
17	Qwen 3.8 27B	max	67	–	–
18	GLM 5.1	max	67	–	–
19	Qwen 3.5 27B	high	61	–	–
20	Gemini 3.1 Flash-Lite	high	59	–	–
21	MiMo V2.5 Pro	xhigh	58	–	–
22	Kimi K2.6	max	57	–	–
23	Qwen 3.5 397B-A17B	max	54	–	–
24	Qwen 3.6 27B	max	53	–	–
25	Gemma 4 26B	xhigh	53	–	–
26	Nemotron 3 Ultra	max	53	–	–
27	Grok 4.3	xhigh	50	–	–
28	DeepSeek V4 Pro	max	47	–	–
29	Gemini 3 Flash	high	46	–	–
30	Qwen 3.6 Max	max	45	–	–
31	Qwen 3.5 35B-A3B	low	41	–	–
32	DeepSeek V4 Flash	max	37	–	–
33	Qwen 3.6 35B-A3B	max	33	–	–
34	Qwen 3.5 9B	low	32	–	–
35	Qwen 3.6 Flash	max	31	–	–
36	DeepSeek V4 Flash 0731	max	31	–	–
37	GPT-5.4 Nano	xhigh	28	–	–
38	GPT-5.4 Mini	xhigh	27	–	–
39	Mistral Small 4	high	16	–	–
40	gpt-oss-120b	high	15	–	–