Best AI models for unsatisfiable constraints

Ranked by accuracy on the meo “unsatisfiable constraints” 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.

#	Model	Lab	accuracy
1	OpenAI: GPT-5.5	openai	100%
2	Google: Gemini 4 Argon	google	100%
3	Anthropic: Claude Opus 5.5	anthropic	100%
4	OpenAI: GPT-5.5 Pro	openai	100%
5	OpenAI: GPT-6.1 Sol	openai	100%
6	Anthropic: Claude Fable 5	anthropic	100%
7	Anthropic: Claude Opus 4.8	anthropic	100%
8	Z.ai: GLM 5.2	z-ai	100%
9	Google: Gemini 3.5 Flash	google	100%
10	Qwen: Qwen3.7 Max	qwen	100%
11	MoonshotAI: Kimi K3	moonshotai	100%
12	OpenAI: GPT-5.4 Nano	openai	100%
13	Qwen: Qwen3.5-Flash	qwen	100%
14	MoonshotAI: Kimi K2.6	moonshotai	100%
15	Qwen: Qwen3 VL 8B Thinking	qwen	100%
16	inclusionAI: Ring-2.6-1T	inclusionai	94%
17	DeepSeek: DeepSeek V4 Flash 0423	deepseek	94%
18	DeepSeek: DeepSeek V4 Pro 0423	deepseek	94%
19	NVIDIA: Nemotron 3 Ultra	nvidia	94%
20	SpaceXAI: Grok 4.7	x-ai	94%
21	Qwen: Qwen3.6 Flash	qwen	94%
22	MiniMax: MiniMax M3	minimax	94%
23	MoonshotAI: Kimi K2.7 Code	moonshotai	94%
24	OpenAI: GPT-5 Mini	openai	94%
25	SpaceXAI: Grok 4.3	x-ai	94%
26	Qwen: Qwen3 VL 30B A3B Instruct	qwen	94%
27	Qwen: Qwen3 VL 235B A22B Instruct	qwen	94%
28	Qwen: Qwen3 VL 30B A3B Thinking	qwen	94%
29	Google: Gemma 4 31B	google	94%
30	Google: Gemini 3.1 Flash Lite Preview	google	94%
31	Qwen: Qwen3.5 Plus 2026-02-15	qwen	94%
32	Google: Gemini 3.1 Pro Preview	google	88%
33	ByteDance Seed: Seed 1.6	bytedance-seed	88%
34	Xiaomi: MiMo-V2.5-Pro	xiaomi	81%
35	Qwen: Qwen3.7 Plus	qwen	81%
36	Xiaomi: MiMo-V2.5	xiaomi	81%
37	Google: Gemini 3.1 Flash Lite	google	81%
38	OpenAI: GPT-5.1-Codex-Mini	openai	75%
39	Qwen: Qwen3 VL 32B Instruct	qwen	75%
40	Nex AGI: Nex-N2-Pro (free)	nex-agi	69%
41	Z.ai: GLM 4.6V	z-ai	69%
42	ByteDance Seed: Seed 1.6 Flash	bytedance-seed	69%
43	Mistral: Mistral Small 4	mistralai	69%
44	Qwen: Qwen3 VL 8B Instruct	qwen	69%
45	Google: Gemma 4 26B A4B	google	69%
46	Baidu: ERNIE 4.5 VL 424B A47B	baidu	63%
47	Owl Alpha	openrouter	56%
48	OpenAI: GPT-4.1 Mini	openai	56%
49	Mistral: Mistral Large 3 2512	mistralai	56%
50	Mistral: Ministral 3 14B 2512	mistralai	56%
51	Qwen: Qwen3.5-27B	qwen	56%
52	NVIDIA: Nemotron 3 Nano Omni (free)	nvidia	56%
53	Z.ai: GLM 5.1	z-ai	50%
54	Google: Gemini 2.5 Flash Lite	google	50%
55	Tencent: Hy3 preview	tencent	50%
56	OpenAI: GPT-5 Nano	openai	50%
57	StepFun: Step 3.7 Flash	stepfun	50%
58	Google: Gemma 3 4B	google	50%
59	Perplexity: Sonar	perplexity	50%
60	Arcee AI: Trinity Large Thinking	arcee-ai	50%
61	Perceptron: Perceptron Mk1	perceptron	44%
62	Mistral: Mistral Small 3.2 24B	mistralai	44%
63	Google: Gemma 3 27B	google	44%
64	Google: Gemma 3 12B	google	44%
65	Mistral: Ministral 3 3B 2512	mistralai	44%
66	Qwen: Qwen3.6 Plus	qwen	44%
67	Mistral: Ministral 3 8B 2512	mistralai	38%
68	Qwen: Qwen3.6 35B A3B	qwen	38%
69	Qwen: Qwen3.5-9B	qwen	38%
70	Meta: Llama 4 Maverick	meta-llama	31%
71	Qwen: Qwen2.5 VL 72B Instruct	qwen	25%
72	MiniMax: MiniMax-01	minimax	25%
73	OpenAI: GPT-4o-mini	openai	25%
74	NVIDIA: Nemotron Nano 12B 2 VL (free)	nvidia	25%
75	Qwen: Qwen3.5-35B-A3B	qwen	25%
76	OpenAI: GPT-4.1 Nano	openai	19%
77	ByteDance Seed: Seed-2.0-Lite	bytedance-seed	19%
78	Anthropic: Claude 3 Haiku	anthropic	19%
79	Meta: Llama 3.2 11B Vision Instruct	meta-llama	19%
80	Meta: Llama 4 Scout	meta-llama	13%
81	Amazon: Nova Lite 1.0	amazon	13%
82	ByteDance Seed: Seed-2.0-Mini	bytedance-seed	0%

← All rankings Methodology & 𝕍 →