🤖 Agentic · llm-stats.com
MCP Atlas
Tool calling over the Model Context Protocol across complex server graphs.
17
Modèles testés
Kimi K3
Modèle phare
84.2%
Meilleur score
Moonshot AI
Fournisseur
| # | Modèle | MCP Atlas |
|---|---|---|
| 1 | Kimi K3 Moonshot AI | 84.2% |
| 2 | Claude Opus 4.8 Anthropic | 82.2% |
| 3 | Hunyuan Hy3 Tencent | 79.1% |
| 4 | Claude Opus 4.7 Anthropic | 77.3% |
| 5 | GLM-5.2 Z.ai | 76.8% |
| 6 | Qwen3.7 Max Alibaba | 76.4% |
| 7 | GPT-5.5 OpenAI | 75.3% |
| 8 | Qwen3.6 Plus Alibaba | 74.1% |
| 9 | DeepSeek V4 Pro DeepSeek | 73.6% |
| 10 | Qwen3.7 Plus Alibaba | 73.2% |
| 11 | Gemini 3.1 Pro Google | 69.2% |
| 12 | DeepSeek V4 Flash DeepSeek | 69% |
| 13 | GPT-5.4 OpenAI | 67.2% |
| 14 | Claude Opus 4.6 Anthropic | 62.7% |
| 15 | Claude Sonnet 4.6 Anthropic | 61.3% |
| 16 | GPT-5.2 OpenAI | 60.6% |
| 17 | Gemini 3 Flash Google | 57.4% |
Données agrégées depuis llm-stats.com. Voir Méthodologie.