SmartBench (Context-Independent): leaderboard
Metric: F1 (%) of anomaly detection (anomalous is the positive class) on the 2,000 context-independent samples (1,000 normal, 1,000 anomalous single-moment device states with environment readings); temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 79.3 | #64 |
| 2 | GPT-5 | 79 | #91 |
| 3 | Gemini 2.5 Pro | 73.5 | #145 |
| 4 | GPT-5 Mini | 72.5 | #176 |
| 5 | DeepSeek R1 | 72 | #245 |
| 6 | Claude Sonnet 4.5 | 68.6 | #138 |
| 7 | Qwen 3 32B | 64.8 | #424 |
| 8 | Claude Sonnet 4 (20250514) | 60.1 | #211 |
| 9 | DeepSeek V3 | 51.3 | #312 |
| 10 | Qwen 3 8B | 46.2 | #667 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=smartbench-context-independent · How It Works · Data refreshed daily, snapshot 2026-10-11.