MCP Atlas public (Kimi K3 100-turn evaluation)
Agents · 2026-07-16
Moonshot's Kimi K3 comparison evaluates all models on MCP-Atlas's 500-task public subset with a 100-turn limit and Gemini 3.1 Pro as judge. This source-specific protocol is distinct from Scale's revised 100-tool-call budget and full 1,000-task leaderboard. Exact task identifiers are not published in the release post.
Top models (higher is better)
| Model | Score |
|---|---|
| Kimi K3 | 84.2 |
| GPT-5.6 Sol | 83.6 |
| Opus 4.8 | 83.6 |
| GPT-5.5 | 82.8 |
| GLM-5.2 | 82.6 |