Compare AI humanizers
Pick two to four entries and inspect them side by side. All numbers come from the same shared source set and measurement profiles. An asterisk marks detector-dependent, provisional values.
| November 2026 | Humanizer 1B · OptiQ 4-bitVerified by HumanizeBenchWe ran this model ourselves on the benchmark source examples and checked its execution records. Verification confirms the tested model’s provenance; it does not guarantee performance or imply that every detector returned a result.mlx-community/humanizer-1B-OptiQ-4bit Open weights |
|---|---|
| Shared score components | |
| Overall score | 17.81 * |
| Panel bypass score | 0.6% * |
| Meaning preservation | 77.1% |
| Raw meaning cosine | 0.9242 |
| Readability | 71.2% |
| Category consistency | 99.0% * |
| Detector coverage | 229/231 |
| Identical outputs | 2/33 |
| Penalty points | −17 |
| Per-detector human-likelihood scores · measured outputs only | |
| GPTZero | 0.0% 31/33 |
| ZeroGPT | 0.0% 33/33 |
| Copyleaks | 3.0% 33/33 |
| Originality.ai | 3.3% 33/33 |
| Pangram | 4.4% 33/33 |
| Grammarly | 28.2% 33/33 |
| QuillBot | 99.3% 33/33 |
| Category scores · before entry-level penalties | |
| Academic essay | 39.3 |
| Application essay | 40.5 |
| Blog post | 29.7 |
| Business email | 35.4 |
| Marketing copy | 35.9 |
| Discussion board | 24.0 |
| News article | 32.9 |
| Evidence | All 33 outputs |