Benchmarks / Agentic
CheatBench
Uses the published cheating-attempt rate; lower is better. Score: cheating rate. Collected by Jake Boggs' benchmark dashboard.
CheatBench
Lower is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Opus 5.5 | max | 11.2% |
| 2 | Muse Spark 1.3 | max | 39.0% |
| 3 | Fable 5.1 | max | 45.8% |
| 4 | GPT-6 Astra | max | 47.4% |
| 5 | GPT-6.1 Sol | max | 49.3% |
| 6 | Kimi K3 | max | 70.0% |
| 7 | DeepSeek V4 Pro | max | 75.1% |
| 8 | Gemini 3.8 Flash | max | 75.2% |
| 9 | Grok 4.7 | max | 78.0% |