Mercor APEX · Open Analysis

#	Model	Effort	Score	Cost / task	Harness
1	Gemini 4	max	82.4%	–	–
2	Sonnet 5.5	max	75.5%	–	–
3	Opus 5.5	max	73.5%	–	–
4	Fable 5.1	max	68.6%	–	–
5	Gemini 3.7 Flash	max	67.8%	–	–
6	Opus 5	max	65.8%	–	–
7	Grok 4.6	max	65.3%	–	–
8	GPT-6 Astra	max	64.7%	–	–
9	Gemini 3.8 Flash	max	64.3%	–	–
10	Fable 5	max	63.6%	–	–
11	Qwen 3.8 Max	max	63.3%	–	–
12	GPT-6.1 Sol	max	60.0%	–	–
13	MiMo V2.6 Pro	high	59.5%	–	–
14	Muse Spark 1.3	max	58.6%	–	–
15	GPT-5.6 Terra	max	58.2%	–	–
16	MiMo V2.6 Flash	high	57.4%	–	–
17	GLM 5.3	max	56.6%	–	–
18	Grok 4.5	max	56.2%	–	–
19	Haiku 5.5	max	56.1%	–	–
20	DeepSeek V4 Flash	max	55.3%	–	–
21	GPT-5.5	xhigh	55.1%	–	–
22	Grok 4.7	max	54.6%	–	–
23	Sonnet 5	max	54.5%	–	–
24	GPT-6 Sol	max	54.3%	–	–
25	GLM 5.3 Flash	max	52.8%	–	–
26	GPT-5.4	xhigh	52.4%	–	–
27	GPT-5.6 Sol	max	51.5%	–	–
28	Kimi K3	max	50.6%	–	–
29	Opus 4.7	max	49.2%	–	–
30	Opus 4.8	max	48.9%	–	–
31	Qwen 3.8 27B	max	47.5%	–	–
32	DeepSeek V4 Pro 0813	max	47.3%	–	–
33	Gemini 3.6 Flash	xhigh	46.9%	–	–
34	Opus 4.6	max	46.3%	–	–
35	GLM 5.2	max	45.2%	–	–
36	GPT-6 Luna	max	44.3%	–	–
37	Sonnet 4.6	max	43.0%	–	–
38	GPT-5.6 Luna	max	43.0%	–	–
39	GLM 5.1	max	40.9%	–	–
40	DeepSeek V4.1 Flash	max	39.5%	–	–
41	MiniMax M3	max	37.7%	–	–
42	Kimi K2.7 Code	max	37.6%	–	–
43	Muse Spark 1.2	max	36.4%	–	–
44	Gemini 3.1 Pro	high	35.3%	–	–
45	Inkling	max	33.8%	–	–
46	Muse Spark 1.1	max	31.8%	–	–
47	Gemini 3.5 Flash-Lite	xhigh	29.3%	–	–
48	Gemini 3.5 Flash	xhigh	27.5%	–	–
49	Qwen 3.5	none	24.9%	–	–
50	Nemotron 3 Ultra 550B A55B	max	22.7%	–	–
51	DeepSeek V3.2	high	21.3%	–	–
52	gpt-oss-120b	high	4.4%	–	–