EQ-Bench 3 · Open Analysis

#	Model	Effort	Score	Cost / task	Harness
1	Fable 5	max	2050	–	–
2	Opus 4.8	max	2030	–	–
3	Opus 4.7	max	1884	–	–
4	Opus 4.6	max	1717	–	–
5	Sonnet 4.6	max	1714	–	–
6	GPT-5.5	xhigh	1577	–	–
7	GLM 5.2	max	1575	–	–
8	DeepSeek V4 Pro	max	1570	–	–
9	GLM 5.1	max	1567	–	–
10	GPT-5.4	xhigh	1563	–	–
11	Kimi K2	medium	1562	–	–
12	Kimi K2.6	max	1561	–	–
13	Gemini 3 Pro	high	1559	–	–
14	GPT-5.2	xhigh	1559	–	–
15	GPT-5.1	high	1551	–	–
16	Opus 4.5	high	1546	–	–
17	Gemini 2.5 Pro	high	1545	–	–
18	Kimi K2.5	max	1545	–	–
19	Gemini 3.1 Pro	high	1538	–	–
20	GLM 5	high	1526	–	–
21	Sonnet 4.5	high	1511	–	–
22	o3	high	1500	–	–
23	DeepSeek V4 Flash	max	1491	–	–
24	Gemma 4 31B IT	high	1434	–	–
25	Gemma 4 26B	xhigh	1431	–	–
26	GLM 4.7	none	1428	–	–
27	Qwen 3.5 397B-A17B	max	1413	–	–
28	Opus 4	medium	1400	–	–
29	GPT-5.3	xhigh	1393	–	–
30	Gemma 4 12B IT	–	1361	–	–
31	GLM 4.5	xhigh	1276	–	–
32	Sonnet 4	high	1243	–	–
33	o4 mini	high	1233	–	–
34	DeepSeek R1	xhigh	1231	–	–
35	Qwen 3 235B-A22B	xhigh	1214	–	–
36	Grok 4.1 Fast	high	1185	–	–
37	GPT-4.1	high	1180	–	–
38	Gemini 2.5 Flash	medium	1165	–	–
39	Grok 3	none	1150	–	–
40	Grok 4	high	1148	–	–
41	Sonnet 3.7	medium	1112	–	–
42	GPT-4.1 Mini	high	1099	–	–
43	Gemma 3 27B IT	medium	1094	–	–
44	GLM 4.7 Flash	none	1083	–	–
45	DeepSeek V3.0324	–	1082	–	–
46	Sonnet 3.5	medium	1081	–	–
47	gpt-oss-120b	high	1076	–	–
48	Mistral Small 3.2	none	1072	–	–
49	Grok 3 Mini	high	1065	–	–
50	GPT-4.5	none	1053	–	–
51	Qwen 3 32B	none	1048	–	–
52	Llama 3.1	max	1043	–	–
53	GPT-4.1 Nano	high	1029	–	–
54	Gemma 3 4B IT	–	977	–	–
55	gpt-oss-20b	high	956	–	–
56	Mistral Large 2411	–	951	–	–
57	Gemini 2.0 Flash	medium	911	–	–
58	Qwen 3 30B-A3B	none	868	–	–
59	Qwen 3 8B	none	866	–	–
60	Grok 4.20	xhigh	852	–	–
61	Qwen 2.5 72B	none	846	–	–
62	Llama 4 Maverick	medium	807	–	–
63	Mistral Small	none	802	–	–
64	Llama 3.1 405b	none	799	–	–
65	Mistral Small 3.1	none	784	–	–
66	Llama 4 Scout	medium	608	–	–
67	GPT-4	high	578	–	–
68	Gemma 2 9B IT	–	554	–	–
69	Llama 3.2 1b	none	200	–	–