Best AI models for logic, math & CS

Ranked by accuracy on the meo “logic, math & CS” 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 2026-10-18.

#	Model	Lab	accuracy
1	Anthropic: Claude Fable 5	anthropic	88%
2	inclusionAI: Ring-2.6-1T	inclusionai	88%
3	Anthropic: Claude Opus 5.5	anthropic	85%
4	Google: Gemini 3.5 Flash	google	85%
5	Qwen: Qwen3 VL 32B Instruct	qwen	85%
6	OpenAI: GPT-5.5	openai	82%
7	Google: Gemini 4 Argon	google	82%
8	OpenAI: GPT-5.5 Pro	openai	82%
9	OpenAI: GPT-6.1 Sol	openai	82%
10	Z.ai: GLM 5.2	z-ai	82%
11	MoonshotAI: Kimi K3	moonshotai	82%
12	DeepSeek: DeepSeek V4 Pro 0423	deepseek	82%
13	DeepSeek: DeepSeek V4 Flash 0423	deepseek	79%
14	Google: Gemini 3.1 Pro Preview	google	79%
15	NVIDIA: Nemotron 3 Ultra	nvidia	79%
16	SpaceXAI: Grok 4.7	x-ai	79%
17	OpenAI: GPT-5.4 Nano	openai	79%
18	MiniMax: MiniMax M3	minimax	79%
19	OpenAI: GPT-5 Mini	openai	79%
20	OpenAI: GPT-5.1-Codex-Mini	openai	79%
21	Xiaomi: MiMo-V2.5	xiaomi	79%
22	Qwen: Qwen3 VL 235B A22B Instruct	qwen	79%
23	Anthropic: Claude Opus 4.8	anthropic	76%
24	Qwen: Qwen3.7 Max	qwen	76%
25	Qwen: Qwen3.6 Flash	qwen	76%
26	MoonshotAI: Kimi K2.6	moonshotai	76%
27	OpenAI: GPT-4.1 Mini	openai	76%
28	Qwen: Qwen3 VL 8B Thinking	qwen	76%
29	Qwen: Qwen3.5 Plus 2026-02-15	qwen	76%
30	ByteDance Seed: Seed 1.6 Flash	bytedance-seed	76%
31	Qwen: Qwen3.5-Flash	qwen	74%
32	MoonshotAI: Kimi K2.7 Code	moonshotai	74%
33	SpaceXAI: Grok 4.3	x-ai	74%
34	Z.ai: GLM 4.6V	z-ai	74%
35	Owl Alpha	openrouter	71%
36	Qwen: Qwen3.7 Plus	qwen	71%
37	Qwen: Qwen3 VL 30B A3B Instruct	qwen	71%
38	Qwen: Qwen3 VL 30B A3B Thinking	qwen	71%
39	Google: Gemma 4 31B	google	71%
40	Mistral: Ministral 3 14B 2512	mistralai	71%
41	Xiaomi: MiMo-V2.5-Pro	xiaomi	68%
42	Baidu: ERNIE 4.5 VL 424B A47B	baidu	68%
43	Mistral: Mistral Small 4	mistralai	68%
44	Qwen: Qwen3 VL 8B Instruct	qwen	65%
45	Mistral: Mistral Small 3.2 24B	mistralai	65%
46	Nex AGI: Nex-N2-Pro (free)	nex-agi	59%
47	ByteDance Seed: Seed 1.6	bytedance-seed	59%
48	Z.ai: GLM 5.1	z-ai	59%
49	Google: Gemini 3.1 Flash Lite Preview	google	59%
50	Google: Gemini 2.5 Flash Lite	google	59%
51	Mistral: Ministral 3 8B 2512	mistralai	59%
52	OpenAI: GPT-5 Nano	openai	59%
53	Mistral: Mistral Large 3 2512	mistralai	56%
54	StepFun: Step 3.7 Flash	stepfun	56%
55	Google: Gemini 3.1 Flash Lite	google	47%
56	Tencent: Hy3 preview	tencent	47%
57	NVIDIA: Nemotron 3 Nano Omni (free)	nvidia	47%
58	Qwen: Qwen3.6 35B A3B	qwen	47%
59	Google: Gemma 4 26B A4B	google	44%
60	Google: Gemma 3 27B	google	44%
61	Perplexity: Sonar	perplexity	35%
62	Meta: Llama 4 Maverick	meta-llama	32%
63	Mistral: Ministral 3 3B 2512	mistralai	32%
64	Google: Gemma 3 12B	google	29%
65	Arcee AI: Trinity Large Thinking	arcee-ai	29%
66	Qwen: Qwen3.5-27B	qwen	26%
67	OpenAI: GPT-4o-mini	openai	26%
68	Google: Gemma 3 4B	google	26%
69	Perceptron: Perceptron Mk1	perceptron	24%
70	Qwen: Qwen2.5 VL 72B Instruct	qwen	24%
71	NVIDIA: Nemotron Nano 12B 2 VL (free)	nvidia	24%
72	OpenAI: GPT-4.1 Nano	openai	21%
73	ByteDance Seed: Seed-2.0-Lite	bytedance-seed	21%
74	MiniMax: MiniMax-01	minimax	18%
75	Meta: Llama 4 Scout	meta-llama	18%
76	Qwen: Qwen3.6 Plus	qwen	18%
77	Amazon: Nova Lite 1.0	amazon	18%
78	Qwen: Qwen3.5-35B-A3B	qwen	18%
79	Anthropic: Claude 3 Haiku	anthropic	15%
80	ByteDance Seed: Seed-2.0-Mini	bytedance-seed	12%
81	Qwen: Qwen3.5-9B	qwen	9%
82	Meta: Llama 3.2 11B Vision Instruct	meta-llama	9%

← All rankings Methodology & 𝕍 →