Skip to main content
Meo

Best AI models for deep agentic chains

Ranked by Effective Value at task depth N=40 โ€” long agentic chains where one error compounds. As depth grows the ranking shifts decisively toward accuracy: fast-but-flawed models collapse while accurate models pull ahead. This is the leaderboard for autonomous, multi-step work. Shown indexed to the top model at N=40 = 100.

Ranking 82 models across the meo-tested roster ยท as of 2026-10-18.

#ModelLab๐• idx (N=40)
1Anthropic: Claude Opus 5.5anthropic100.0
2OpenAI: GPT-5.5openai58.7
3Google: Gemini 4 Argongoogle21.9
4OpenAI: GPT-6.1 Solopenai16.7
5Anthropic: Claude Fable 5anthropic6.0
6OpenAI: GPT-5.5 Proopenai3.5
7inclusionAI: Ring-2.6-1Tinclusionai0.49
8Anthropic: Claude Opus 4.8anthropic0.46
9Google: Gemini 3.5 Flashgoogle0.10
10MoonshotAI: Kimi K3moonshotai0.015
11Z.ai: GLM 5.2z-ai0.013
12Google: Gemini 3.1 Pro Previewgoogle0.009
13Qwen: Qwen3.7 Maxqwen0.004
14DeepSeek: DeepSeek V4 Flash 0423deepseek0.004
15DeepSeek: DeepSeek V4 Pro 0423deepseek0.002
16NVIDIA: Nemotron 3 Ultranvidia<0.001
17Owl Alphaopenrouter<0.001
18OpenAI: GPT-5.4 Nanoopenai<0.001
19SpaceXAI: Grok 4.7x-ai<0.001
20Xiaomi: MiMo-V2.5-Proxiaomi<0.001
21SpaceXAI: Grok 4.3x-ai<0.001
22OpenAI: GPT-5.1-Codex-Miniopenai<0.001
23Qwen: Qwen3.6 Flashqwen<0.001
24Qwen: Qwen3.5-Flashqwen<0.001
25OpenAI: GPT-5 Miniopenai<0.001
26Xiaomi: MiMo-V2.5xiaomi<0.001
27MiniMax: MiniMax M3minimax<0.001
28Qwen: Qwen3 VL 32B Instructqwen<0.001
29Nex AGI: Nex-N2-Pro (free)nex-agi<0.001
30Qwen: Qwen3 VL 30B A3B Instructqwen<0.001
31MoonshotAI: Kimi K2.7 Codemoonshotai<0.001
32Z.ai: GLM 4.6Vz-ai<0.001
33Qwen: Qwen3.7 Plusqwen<0.001
34MoonshotAI: Kimi K2.6moonshotai<0.001
35OpenAI: GPT-4.1 Miniopenai<0.001
36Google: Gemini 3.1 Flash Lite Previewgoogle<0.001
37Qwen: Qwen3 VL 235B A22B Instructqwen<0.001
38ByteDance Seed: Seed 1.6bytedance-seed<0.001
39Qwen: Qwen3 VL 8B Thinkingqwen<0.001
40Z.ai: GLM 5.1z-ai<0.001
41Baidu: ERNIE 4.5 VL 424B A47B baidu<0.001
42Qwen: Qwen3 VL 30B A3B Thinkingqwen<0.001
43Mistral: Mistral Large 3 2512mistralai<0.001
44Google: Gemma 4 31Bgoogle<0.001
45Qwen: Qwen3.5 Plus 2026-02-15qwen<0.001
46Google: Gemini 2.5 Flash Litegoogle<0.001
47ByteDance Seed: Seed 1.6 Flashbytedance-seed<0.001
48Mistral: Mistral Small 4mistralai<0.001
49Google: Gemini 3.1 Flash Litegoogle<0.001
50Perceptron: Perceptron Mk1perceptron<0.001
51Mistral: Ministral 3 14B 2512mistralai<0.001
52Qwen: Qwen3 VL 8B Instructqwen<0.001
53Mistral: Ministral 3 8B 2512mistralai<0.001
54OpenAI: GPT-5 Nanoopenai<0.001
55Mistral: Mistral Small 3.2 24Bmistralai<0.001
56Google: Gemma 4 26B A4B google<0.001
57Tencent: Hy3 previewtencent<0.001
58Meta: Llama 4 Maverickmeta-llama<0.001
59OpenAI: GPT-4.1 Nanoopenai<0.001
60Google: Gemma 3 27Bgoogle<0.001
61Qwen: Qwen2.5 VL 72B Instructqwen<0.001
62NVIDIA: Nemotron 3 Nano Omni (free)nvidia<0.001
63Google: Gemma 3 12Bgoogle<0.001
64StepFun: Step 3.7 Flashstepfun<0.001
65OpenAI: GPT-4o-miniopenai<0.001
66MiniMax: MiniMax-01minimax<0.001
67Meta: Llama 4 Scoutmeta-llama<0.001
68Mistral: Ministral 3 3B 2512mistralai<0.001
69Qwen: Qwen3.6 35B A3Bqwen<0.001
70Qwen: Qwen3.5-27Bqwen<0.001
71Google: Gemma 3 4Bgoogle<0.001
72Perplexity: Sonarperplexity<0.001
73ByteDance Seed: Seed-2.0-Litebytedance-seed<0.001
74Anthropic: Claude 3 Haikuanthropic<0.001
75NVIDIA: Nemotron Nano 12B 2 VL (free)nvidia<0.001
76Amazon: Nova Lite 1.0amazon<0.001
77Qwen: Qwen3.6 Plusqwen<0.001
78Arcee AI: Trinity Large Thinkingarcee-ai<0.001
79Qwen: Qwen3.5-9Bqwen<0.001
80Qwen: Qwen3.5-35B-A3Bqwen<0.001
81Meta: Llama 3.2 11B Vision Instructmeta-llama<0.001
82ByteDance Seed: Seed-2.0-Minibytedance-seed<0.001

โ† All rankingsMethodology & ๐• โ†’