Incompressible Knowledge Probes (IKP): leaderboard
Metric: Mean tier accuracy (%). Source: 01.me. 133 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) (Medium) | 86.19 |
| 2 | Gemini 3.1 Pro (Preview) | 85.6 |
| 3 | Gemini 3 Flash (Preview) | 85.06 |
| 4 | GPT-5.5 | 82.56 |
| 5 | GPT-5.5 (Medium) | 82.56 |
| 6 | Claude Fable 5 | 80.49 |
| 7 | Gemini 2.5 Pro (Preview) | 79.48 |
| 8 | Gemini 2.5 Pro (Medium) | 78.87 |
| 9 | GPT-5.4 Pro (xHigh) | 77.84 |
| 10 | GPT-4.1 | 77.18 |
| 11 | GPT-5 Pro | 76.98 |
| 12 | O3 | 76.91 |
| 13 | Gemini 3.1 Flash Lite (Preview) | 76.8 |
| 14 | GPT-5 (Medium) | 76.52 |
| 15 | GPT-5 | 75.82 |
Interactive version: theaggregate.ai/benchmark?slug=incompressible-knowledge-probes-ikp · How It Works · Data refreshed daily, snapshot 2026-09-19.