November 2026 · same 50 inputs
StealthGPT Super vs Grubby
November 2026 · methodology v2 · provisional automated quality. 15 humanizers, 50 identical inputs each, five detectors and eight quality metrics. Read the scoring rules and post-run changes.
| Overall rank | Humanizer | Overall average | Overall median | Weighted bypass | Quality | |
|---|---|---|---|---|---|---|
| 2 | StealthGPT SuperModel: Super · Mode: Quality; rephrase enabled | 70.34 | 74.05 | 69.00% | 73.48 | 1.08 |
| 9 | GrubbyModel: Not exposed · Mode: Academic | 28.36 | 26.85 | 10.40% | 70.28 | 1.19 |
StealthGPT Super: detector breakdown
| Detector | Model | Weight | Passes | Pass rate | Native human-score average |
|---|---|---|---|---|---|
| Pangram 4.1 | 4.1 | 40% | 44/50 | 88.00% | 95.19 |
| GPTZero 5.0b | 5.0b / 2026-11-04-base | 30% | 30/50 | 60.00% | 52.21 |
| Originality.ai 5.0 | model-5.0 | 20% | 19/50 | 38.00% | 44.38 |
| Winston AI 5.2 | 5.2 | 5% | 48/50 | 96.00% | 97.85 |
| ZeroGPT | Vendor-managed; no public version identifier | 5% | 34/50 | 68.00% | 83.24 |
Grubby: detector breakdown
| Detector | Model | Weight | Passes | Pass rate | Native human-score average |
|---|---|---|---|---|---|
| Pangram 4.1 | 4.1 | 40% | 1/50 | 2.00% | 8.68 |
| GPTZero 5.0b | 5.0b / 2026-11-04-base | 30% | 4/50 | 8.00% | 6.87 |
| Originality.ai 5.0 | model-5.0 | 20% | 0/50 | 0.00% | 7.42 |
| Winston AI 5.2 | 5.2 | 5% | 50/50 | 100.00% | 99.79 |
| ZeroGPT | Vendor-managed; no public version identifier | 5% | 22/50 | 44.00% | 70.99 |
Overall weights humanization 70% and quality 30%. Rankings use automated quality scores with unresolved evidence flags; neither a detector pass nor a small score gap proves a universal advantage.
Per-output scores CSV · Frozen inputs and prompts CSV · Scoring manifest · Verification bundle · GitHub report · Earlier editions