APEX-Agents-AA: leaderboard
Artificial Analysis implementation of APEX-Agents using the Stirrup agent harness for long-horizon, cross-application professional-services tasks.
Metric: Pass@1 (%). Source: artificialanalysis.ai. Status: years away from saturation. 30 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.5 Flash (High) | 47.05 |
| 2 | Kimi K3 (Max) | 41.3 |
| 3 | GPT-5.6 Terra (Max) | 38.94 |
| 4 | GPT-5.5 (xHigh) | 37.68 |
| 5 | GPT-5.6 Luna (Max) | 35.84 |
| 6 | GLM-5.2 (Max) | 33.7 |
| 7 | GPT-5.4 (xHigh) | 33.26 |
| 8 | Claude Opus 4.6 (Adaptive Reasoning, Max Effort) | 33.04 |
| 9 | Gemini 3.1 Pro (Preview) | 32.01 |
| 10 | Kimi K2.6 | 28.47 |
| 11 | GPT-5.4 Mini (xHigh) | 28.17 |
| 12 | Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort) | 28.02 |
| 13 | Gemini 3 Flash (Preview) (Reasoning) | 27.73 |
| 14 | GPT-5.4 Nano (xHigh) | 24.93 |
| 15 | DeepSeek V4 Pro (Reasoning, Max Effort) | 24.26 |
Interactive version: theaggregate.ai/benchmark?slug=apex-agents-aa · How It Works · Data refreshed daily, snapshot 2026-09-05.