Skip to main content

Best AI models for multi-step state tracking

Ranked by accuracy on the meo “multi-step state tracking” 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.

#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%
7Z.ai: GLM 5.2z-ai100%
8MoonshotAI: Kimi K3moonshotai100%
9Google: Gemini 3.1 Pro Previewgoogle100%
10Owl Alphaopenrouter100%
11SpaceXAI: Grok 4.7x-ai100%
12Qwen: Qwen3.5-Flashqwen100%
13MoonshotAI: Kimi K2.7 Codemoonshotai100%
14Qwen: Qwen3.7 Plusqwen100%
15SpaceXAI: Grok 4.3x-ai100%
16Qwen: Qwen3 VL 30B A3B Instructqwen100%
17Nex AGI: Nex-N2-Pro (free)nex-agi100%
18Z.ai: GLM 4.6Vz-ai100%
19ByteDance Seed: Seed 1.6bytedance-seed100%
20Qwen: Qwen3 VL 235B A22B Instructqwen100%
21MiniMax: MiniMax-01minimax100%
22NVIDIA: Nemotron Nano 12B 2 VL (free)nvidia100%
23Anthropic: Claude Opus 4.8anthropic0%
24inclusionAI: Ring-2.6-1Tinclusionai0%
25Google: Gemini 3.5 Flashgoogle0%
26DeepSeek: DeepSeek V4 Flash 0423deepseek0%
27Qwen: Qwen3.7 Maxqwen0%
28DeepSeek: DeepSeek V4 Pro 0423deepseek0%
29NVIDIA: Nemotron 3 Ultranvidia0%
30Xiaomi: MiMo-V2.5-Proxiaomi0%
31OpenAI: GPT-5.4 Nanoopenai0%
32Qwen: Qwen3.6 Flashqwen0%
33MiniMax: MiniMax M3minimax0%
34MoonshotAI: Kimi K2.6moonshotai0%
35OpenAI: GPT-5 Miniopenai0%
36OpenAI: GPT-5.1-Codex-Miniopenai0%
37Xiaomi: MiMo-V2.5xiaomi0%
38Qwen: Qwen3 VL 32B Instructqwen0%
39Qwen: Qwen3 VL 30B A3B Thinkingqwen0%
40Google: Gemma 4 31Bgoogle0%
41OpenAI: GPT-4.1 Miniopenai0%
42Z.ai: GLM 5.1z-ai0%
43Baidu: ERNIE 4.5 VL 424B A47B baidu0%
44Qwen: Qwen3 VL 8B Thinkingqwen0%
45Google: Gemini 3.1 Flash Lite Previewgoogle0%
46Qwen: Qwen3.5 Plus 2026-02-15qwen0%
47Mistral: Mistral Large 3 2512mistralai0%
48ByteDance Seed: Seed 1.6 Flashbytedance-seed0%
49Mistral: Mistral Small 4mistralai0%
50Perceptron: Perceptron Mk1perceptron0%
51Google: Gemini 2.5 Flash Litegoogle0%
52Qwen: Qwen3 VL 8B Instructqwen0%
53Google: Gemini 3.1 Flash Litegoogle0%
54Mistral: Ministral 3 14B 2512mistralai0%
55Google: Gemma 4 26B A4B google0%
56Tencent: Hy3 previewtencent0%
57Mistral: Ministral 3 8B 2512mistralai0%
58OpenAI: GPT-5 Nanoopenai0%
59Mistral: Mistral Small 3.2 24Bmistralai0%
60Meta: Llama 4 Maverickmeta-llama0%
61Google: Gemma 3 27Bgoogle0%
62Qwen: Qwen2.5 VL 72B Instructqwen0%
63StepFun: Step 3.7 Flashstepfun0%
64Qwen: Qwen3.5-27Bqwen0%
65Google: Gemma 3 12Bgoogle0%
66Meta: Llama 4 Scoutmeta-llama0%
67NVIDIA: Nemotron 3 Nano Omni (free)nvidia0%
68OpenAI: GPT-4.1 Nanoopenai0%
69ByteDance Seed: Seed-2.0-Litebytedance-seed0%
70OpenAI: GPT-4o-miniopenai0%
71Qwen: Qwen3.6 35B A3Bqwen0%
72Google: Gemma 3 4Bgoogle0%
73Mistral: Ministral 3 3B 2512mistralai0%
74Perplexity: Sonarperplexity0%
75Qwen: Qwen3.6 Plusqwen0%
76Anthropic: Claude 3 Haikuanthropic0%
77Arcee AI: Trinity Large Thinkingarcee-ai0%
78Qwen: Qwen3.5-9Bqwen0%
79Amazon: Nova Lite 1.0amazon0%
80Qwen: Qwen3.5-35B-A3Bqwen0%
81ByteDance Seed: Seed-2.0-Minibytedance-seed0%
82Meta: Llama 3.2 11B Vision Instructmeta-llama0%

← All rankingsMethodology & 𝕍 →