SimpleBench

Trick questions where humans score ~84%: spatio-temporal reasoning and social intelligence.

Published by

SimpleBenchdata via Epoch AI Benchmarking Hub

Category

Reasoning

Index weight

1.0

Models

85

Data as of

9 Sept 2026

CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers.

1. 1Claude Fable 581.9%
2. 2Claude Opus 580.6%
3. 3Gemini 3.1 Pro79.6%
4. 4GPT-5.5 Pro76.9%
5. 5GPT-5.5 Pro76.9%
6. 6Gemini 3.5 Flash76.7%
7. 7Gemini 3 Pro76.4%
8. 8Grok 4.675.9%
9. 9Muse Spark 1.274.5%
10. 10GPT-5.4 Pro74.1%
11. 11GPT-5.6 Sol71.7%
12. 12GPT-5.6 Sol71.7%
13. 13Qwen3 770.4%
14. 14Grok 4.570.0%
15. 15GPT-5.569.0%

SearchAll providersAlibabaAnthropicCohereDeepSeekGoogleMetaMiniMaxMistral AIMoonshot AIOpenAIxAIZhipu AIOpen weights85 of 85

#	Model	Score ↓	Run ↕	Index ↕	Published ↕
1	Claude Fable 5Anthropic	81.9%	–	65.8	–
2	Claude Opus 5Anthropic	80.6%	–	70.0	–
3	Gemini 3.1 ProGoogle	79.6%	–	63.9	–
4	GPT-5.5 ProOpenAI	76.9%	–	–	–
5	GPT-5.5 ProOpenAI	76.9%	–	–	–
6	Gemini 3.5 FlashGoogle	76.7%	–	60.7	–
7	Gemini 3 ProGoogle	76.4%	–	61.6	–
8	Grok 4.6xAI	75.9%	grok-4.6_unknown	63.6	–
9	Muse Spark 1.2Meta	74.5%	–	62.6	–
10	GPT-5.4 ProOpenAI	74.1%	–	64.2	–
11	GPT-5.6 SolOpenAI	71.7%	–	–	–
12	GPT-5.6 SolOpenAI	71.7%	–	–	–
13	Qwen3 7Alibaba	70.4%	–	58.9	–
14	Grok 4.5xAI	70.0%	grok-4.5_unknown	63.4	–
15	GPT-5.5OpenAI	69.0%	gpt-5.5_unknown	66.6	–
16	Claude Opus 4.6Anthropic	67.6%	–	63.1	–
17	GPT-5.6 SolOpenAI	64.8%	–	67.5	–
18	GPT-5.6 SolOpenAI	64.8%	–	65.3	–
19	Claude Opus 4.8Anthropic	64.8%	–	64.7	–
20	Qwen3 6Alibaba	63.0%	–	59.5	–
21	Gemini 2.5 ProGoogle	62.4%	–	54.1	–
22	Claude Opus 4.5Anthropic	62.0%	–	58.0	–
23	Claude Opus 4.7Anthropic	61.7%	–	65.5	–
24	GPT-5 ProOpenAI	61.6%	–	60.7	–
25	GPT-5 ProOpenAI	61.6%	–	–	–
26	DeepSeek V4 FlashDeepSeekopen	61.1%	0731	59.1	–
27	Gemini 3 FlashGoogle	61.1%	–	56.6	–
28	Kimi K3Moonshot AIopen	60.7%	–	62.7	–
29	Claude Sonnet 5Anthropic	60.6%	–	61.2	–
30	Grok 4xAI	60.5%	0709	57.2	–
31	Claude Opus 4.1Anthropic	60.0%	–	52.6	–
32	GLM-5.2Zhipu AIopen	58.8%	glm-5.2_unknown	57.4	–
33	Claude Opus 4Anthropic	58.8%	–	50.7	–
34	Kimi K2.7 CodeMoonshot AIopen	57.9%	–	56.2	–
35	GPT-5.2 ProOpenAI	57.4%	–	–	–
36	GPT-5OpenAI	56.7%	–	54.5	–
37	Grok 4.1xAI	56.0%	–	55.9	–
38	Grok 4.1xAI	56.0%	–	47.9	–
39	GLM-5.1Zhipu AIopen	55.1%	glm-5.1	60.1	–
40	Claude Sonnet 4.5Anthropic	54.3%	–	52.3	–
41	GLM-5Zhipu AIopen	53.2%	glm-5	59.0	–
42	GPT-5.1OpenAI	53.2%	–	53.8	–
43	o3OpenAI	53.1%	–	52.9	–
44	DeepSeek V3.2DeepSeekopen	52.6%	–	54.7	–
45	GPT-5.6 TerraOpenAI	48.9%	–	64.2	–
46	GPT-5.6 TerraOpenAI	48.9%	–	62.3	–
47	GLM-4.7Zhipu AIopen	47.7%	–	54.9	–
48	GPT-5.6 LunaOpenAI	46.8%	–	59.3	–
49	GPT-5.6 LunaOpenAI	46.8%	–	56.6	–
50	Kimi K2.5Moonshot AIopen	46.8%	kimi-k2.5	56.1	–
51	Claude 3.7 SonnetAnthropic	46.4%	–	49.8	–
52	GPT-5.2OpenAI	45.8%	–	61.5	–
53	MiniMax-M3MiniMaxopen	45.8%	MiniMax-M3	56.9	–
54	GPT-5.2OpenAI	45.8%	–	55.5	–
55	Claude Sonnet 4Anthropic	45.5%	–	49.5	–
56	o1OpenAI	41.7%	–	53.2	–
57	Claude 3.5 SonnetAnthropic	41.4%	–	45.1	–
58	Gemini 2.5 FlashGoogle	41.2%	–	48.8	–
59	DeepSeek R1DeepSeekopen	40.8%	0528	48.9	–
60	o1OpenAI	40.1%	–	44.3	–

Show 25 more