APEX-Agents-AA: leaderboard
An agent completes a long-horizon task across files and tools in investment banking, management consulting or corporate law. Artificial Analysis runs 452 of the public tasks on its Stirrup harness, and an LLM grader passes a task only if every binary rubric criterion is met. Mirrored via benchmarklist.com with a few vendor launch-post rows; higher is better. Source: artificialanalysis.ai/evaluations/apex-agents-aa.
Metric: Pass@1 (verified). Source: benchmarklist.com. Saturation forecast: Around October 2027. 57 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 (Max) | 59.2 | #16 (Claude Fable 5) |
| 2 | Grok 4.6 (High) | 57.5 | #26 (Grok 4.6) |
| 3 | GPT-5.6 Sol (Max) | 56.7 | #19 (GPT-5.6 Sol) |
| 4 | Gemini 3.5 Flash (High) | 47.1 | #60 (Gemini 3.5 Flash) |
| 5 | Grok 4.5 (High) | 47.1 | #48 (Grok 4.5) |
| 6 | Kimi K3 | 41.3 | #30 |
| 7 | GPT-5.6 Terra | 38.94 | #41 |
| 8 | GPT-5.5 (xHigh) | 37.7 | #31 (GPT-5.5) |
| 9 | GPT-5.5 | 37.68 | #31 |
| 10 | GPT-5.6 Luna | 35.84 | #68 |
| 11 | GPT-5.4 (xHigh) | 33.3 | #79 (GPT-5.4) |
| 12 | GPT-5.4 | 33.26 | #79 |
| 13 | Claude Opus 4.6 (Adaptive Reasoning, Max Effort) | 33 | #63 (Claude Opus 4.6) |
| 14 | Gemini 3.1 Pro (Preview) | 32 | #58 |
| 15 | Kimi K2.6 | 28.5 | #104 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=apex-agents-aa · How It Works · Data refreshed daily, snapshot 2026-10-20.