Skip to main content

Best AI models for theory of mind

Ranked by accuracy on the meo “theory of mind” domain — one of eleven un-leaked, contamination-resistant domains. Per-domain strength reveals where a model genuinely reasons versus where it pattern-matches; the leaders differ sharply by domain.

Ranking 82 models across the meo-tested roster · as of 2026-10-19.

#ModelLabaccuracy
1OpenAI: GPT-5.5openai100%
2Google: Gemini 4 Argongoogle100%
3Anthropic: Claude Opus 5.5anthropic100%
4OpenAI: GPT-5.5 Proopenai100%
5OpenAI: GPT-6.1 Solopenai100%
6Anthropic: Claude Fable 5anthropic100%
7Anthropic: Claude Opus 4.8anthropic100%
8inclusionAI: Ring-2.6-1Tinclusionai100%
9Z.ai: GLM 5.2z-ai100%
10Google: Gemini 3.5 Flashgoogle100%
11DeepSeek: DeepSeek V4 Flash 0423deepseek100%
12Qwen: Qwen3.7 Maxqwen100%
13MoonshotAI: Kimi K3moonshotai100%
14DeepSeek: DeepSeek V4 Pro 0423deepseek100%
15Google: Gemini 3.1 Pro Previewgoogle100%
16NVIDIA: Nemotron 3 Ultranvidia100%
17Owl Alphaopenrouter100%
18SpaceXAI: Grok 4.7x-ai100%
19Xiaomi: MiMo-V2.5-Proxiaomi100%
20OpenAI: GPT-5.4 Nanoopenai100%
21Qwen: Qwen3.6 Flashqwen100%
22Qwen: Qwen3.5-Flashqwen100%
23MiniMax: MiniMax M3minimax100%
24MoonshotAI: Kimi K2.6moonshotai100%
25MoonshotAI: Kimi K2.7 Codemoonshotai100%
26OpenAI: GPT-5 Miniopenai100%
27Qwen: Qwen3.7 Plusqwen100%
28OpenAI: GPT-5.1-Codex-Miniopenai100%
29SpaceXAI: Grok 4.3x-ai100%
30Xiaomi: MiMo-V2.5xiaomi100%
31Qwen: Qwen3 VL 32B Instructqwen100%
32Nex AGI: Nex-N2-Pro (free)nex-agi100%
33ByteDance Seed: Seed 1.6bytedance-seed100%
34Qwen: Qwen3 VL 235B A22B Instructqwen100%
35Qwen: Qwen3 VL 30B A3B Thinkingqwen100%
36Google: Gemma 4 31Bgoogle100%
37Z.ai: GLM 5.1z-ai100%
38Baidu: ERNIE 4.5 VL 424B A47B baidu100%
39Google: Gemini 3.1 Flash Lite Previewgoogle100%
40Qwen: Qwen3.5 Plus 2026-02-15qwen100%
41Mistral: Mistral Large 3 2512mistralai100%
42Google: Gemini 2.5 Flash Litegoogle100%
43Google: Gemini 3.1 Flash Litegoogle100%
44Google: Gemma 4 26B A4B google100%
45Meta: Llama 4 Maverickmeta-llama100%
46Qwen: Qwen2.5 VL 72B Instructqwen100%
47StepFun: Step 3.7 Flashstepfun100%
48Qwen: Qwen3.5-27Bqwen100%
49Qwen: Qwen3.6 35B A3Bqwen100%
50Qwen: Qwen3.6 Plusqwen100%
51Qwen: Qwen3 VL 30B A3B Instructqwen67%
52Z.ai: GLM 4.6Vz-ai67%
53OpenAI: GPT-4.1 Miniopenai67%
54ByteDance Seed: Seed 1.6 Flashbytedance-seed67%
55Perceptron: Perceptron Mk1perceptron67%
56Mistral: Ministral 3 14B 2512mistralai67%
57Tencent: Hy3 previewtencent67%
58OpenAI: GPT-5 Nanoopenai67%
59Mistral: Mistral Small 3.2 24Bmistralai67%
60Google: Gemma 3 27Bgoogle67%
61MiniMax: MiniMax-01minimax67%
62Google: Gemma 3 12Bgoogle67%
63NVIDIA: Nemotron 3 Nano Omni (free)nvidia67%
64ByteDance Seed: Seed-2.0-Litebytedance-seed67%
65OpenAI: GPT-4o-miniopenai67%
66Qwen: Qwen3.5-9Bqwen67%
67Qwen: Qwen3.5-35B-A3Bqwen67%
68ByteDance Seed: Seed-2.0-Minibytedance-seed67%
69Meta: Llama 3.2 11B Vision Instructmeta-llama67%
70Qwen: Qwen3 VL 8B Thinkingqwen33%
71Mistral: Mistral Small 4mistralai33%
72Qwen: Qwen3 VL 8B Instructqwen33%
73Meta: Llama 4 Scoutmeta-llama33%
74Arcee AI: Trinity Large Thinkingarcee-ai33%
75Amazon: Nova Lite 1.0amazon33%
76Mistral: Ministral 3 8B 2512mistralai0%
77OpenAI: GPT-4.1 Nanoopenai0%
78Google: Gemma 3 4Bgoogle0%
79Mistral: Ministral 3 3B 2512mistralai0%
80Perplexity: Sonarperplexity0%
81NVIDIA: Nemotron Nano 12B 2 VL (free)nvidia0%
82Anthropic: Claude 3 Haikuanthropic0%

← All rankingsMethodology & 𝕍 →