AgencyBench V2 - Overall: leaderboard
Metric: Average Task Score (%). Source: agencybench.opensii.ai. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 | 56.5 |
| 2 | Claude Opus 4.5 | 47.7 |
| 3 | Gemini 3 Pro | 46.9 |
| 4 | Claude Sonnet 4.5 | 46.4 |
| 5 | Grok 4.1 Fast | 44.3 |
| 6 | GLM-4.6 | 38.6 |
| 7 | Kimi K2 (Thinking) | 34.2 |
| 8 | DeepSeek V3.2 | 28.6 |
| 9 | Qwen 3 235B A22B (Thinking) | 27 |
Interactive version: theaggregate.ai/benchmark?slug=agencybench-v2-overall · How It Works · Data refreshed daily, snapshot 2026-09-19.