MCP Atlas
- Moonshot AI
Results
| Model | Score | Source | Human-verified |
|---|---|---|---|
| Kimi K316 Jul 2026 · Moonshot AI | 84.2% | Kimi K3: Open Frontier IntelligenceReasoning effort max; 500-task public subset, 100-turn limit, Gemini 3.1 Pro as the judge (table note) | Yes |
Benchmark
labs.scale.comHigher is better
Scale AIIntroduced 19 Sep 2025