🏆 Top 10 MOE — Artificial Analysis

Fusion Small (4B-40B) + Medium (40B-150B) — tri global par Intelligence Index

Scrapé le samedi 8 août 2026 à 02:02:42 UTC

10
Modèles MoE
Ling 3.0 Flash
Meilleur MoE
37.82
Score Top
#6
Créateurs

Classement unifié Top 10 Mixture of Experts

activeParams < passiveParams × 0.5
#ModèleCréateurIntelligenceParamètresContexte
1 Ling 3.0 Flash ByteDance 37.82 Active: 5.1B / Total: 118.9B 262K
2 Qwen3.5 122B A10B Alibaba 32.85 Active: 10.0B / Total: 115.0B 262K
3 Qwen3.6 35B A3B Alibaba 32.13 Active: 3.0B / Total: 33.0B 262K
4 G9v3-39A5B – 31.39 Active: 5.0B / Total: 34.0B 131K
5 Gemma 4 26B A4B Google 26.07 Active: 3.8B / Total: 21.4B 256K
6 Nemotron 3 Super NVIDIA 25.67 Active: 12.7B / Total: 107.9B 1.0M
7 Qwen3.5 35B A3B Alibaba 24.32 Active: 3.0B / Total: 33.0B 262K
8 gpt-oss-120b (high) Google 24.13 Active: 5.1B / Total: 111.9B 131K
9 Qwen3 Coder Next Alibaba 21.29 Active: 3.0B / Total: 76.7B 256K
10 Mistral Small 4 Mistral AI 19.71 Active: 6.5B / Total: 112.5B 256K