Benchmarks/SimpleBench

SimpleBench

SimpleBench — tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.

reasoning

Models Tested

75.5

Top Score

27.7

Average Score

Rankings

#	Model	Score
1	Gemini 3.1 Pro PreviewGoogle DeepMind	75.5
2	Gemini 3 ProGoogle	71.7
3	GPT-5.4 ProOpenAI	68.9
4	Claude Opus 4.6Anthropic	61.1
5	Gemini 2.5 ProGoogle DeepMind	54.9
6	Claude Opus 4.5Anthropic	54.4
7	GPT-5 ProOpenAI	53.9
8	Gemini 3 Flash PreviewGoogle DeepMind	53.3
9	Grok 4xAI	52.6
10	Claude Opus 4.1Anthropic	52.0
11	Claude Opus 4Anthropic	50.6
12	GPT-5 ChatOpenAI	48.0
13	GPT-5OpenAI	48.0
14	Claude Sonnet 4.5Anthropic	45.2
15	ZA GLM 5 Turboz-ai	43.8
16	ZA GLM 5z-ai	43.8
17	GPT-5.1OpenAI	43.8
18	GPT-5.1 ChatOpenAI	43.8
19	o3OpenAI	43.7
20	ZA GLM 4.7z-ai	37.2
21	M Kimi K2.5moonshotai	36.2
22	Claude 3.7 SonnetAnthropic	35.7
23	Claude 3.7 Sonnet (thinking)Anthropic	35.7
24	GPT-5.2 ChatOpenAI	35.0
25	GPT-5.2OpenAI	35.0
26	Claude Sonnet 4Anthropic	34.6
27	o1-previewOpenAI	30.0
28	DeepSeek-R1 (May 2025)DeepSeek	29.0
29	o1OpenAI	28.1
30	DeepSeek V3.1DeepSeek	28.0
31	o4 MiniOpenAI	26.4
32	Grok 3xAI	23.3
33	Grok 3 BetaxAI	23.3
34	GPT-4.5OpenAI	21.4
35	Gemini 2.0 FlashGoogle DeepMind	17.3
36	Qwen3 235B A22BAlibaba Qwen	17.2
37	DeepSeek-R1DeepSeek	17.1
38	Gemini 2.0 Flash Thinking (Jan 2025)Google	16.8
39	Llama 4 MaverickMeta	13.2
40	Claude 3.5 SonnetAnthropic	13.0
41	Gemini 1.5 Pro (Sept 2024)Google	12.5
42	GPT-4.1OpenAI	12.4
43	M Kimi K2 0711moonshotai	11.6
44	GPT-4 Turbo PreviewOpenAI	10.1
45	Claude 3 OpusAnthropic	8.2
46	Llama 3.1-405BMeta	7.6
47	o3 MiniOpenAI	7.4
48	Grok-2 (Dec 2024)xAI	7.2
49	Mistral Large 2407Mistral AI	7.0
50	gpt-oss-120b (free)OpenAI	6.5
51	gpt-oss-120bOpenAI	6.5
52	gpt-oss-20b (free)OpenAI	6.5
53	gpt-oss-20bOpenAI	6.5
54	Llama 3.3 70B Instruct (free)Meta	3.9
55	Llama 3.3 70B InstructMeta	3.9
56	DeepSeek V3DeepSeek	2.7
57	o1-miniOpenAI	1.7
58	GPT-4o (2024-08-06)OpenAI	1.4
59	GPT-4o (2024-05-13)OpenAI	1.4
60	GPT-4oOpenAI	1.4
61	GPT-4o (extended)OpenAI	1.4

Related Benchmarks