Humanizer CMF vs. Rephrasy
How does the leading open-weight entry in the November snapshot compare with the top commercial entry on readability?
Across 33 matched rewrites with numeric readability scores, Humanizer · CMF scored higher on 22, Rephrasy on 9, and 2 tied. Their November composite scores were 45.23 and 55.34 out of 100, respectively.
Results on the same source texts
Each row compares both entries on the same source and includes pairs with two numeric scores. Meaning uses recorded embedding cosine. Detector scores reflect reported human-likelihood.
| Metric | Humanizer · CMF higher | Rephrasy higher | Tied | Measured pairs |
|---|---|---|---|---|
| Readability | 22 | 9 | 2 | 33 / 33 |
| Meaning preservation | 30 | 3 | 0 | 33 / 33 |
| GPTZero | 0 | 22 | 11 | 33 / 33 |
| ZeroGPT | 3 | 23 | 7 | 33 / 33 |
| Copyleaks | 0 | 15 | 18 | 33 / 33 |
| Originality.ai | 4 | 28 | 1 | 33 / 33 |
| Pangram | 6 | 26 | 1 | 33 / 33 |
| Grammarly | 7 | 15 | 11 | 33 / 33 |
| QuillBot | 4 | 17 | 12 | 33 / 33 |
Read the matched rewrites
These examples show the largest measured gaps in readability, with one example per writing category.
- Blog post ? S16
Humanizer · CMF rewrite (80.0%) ? Rephrasy rewrite (46.7%) - Academic essay ? S23
Humanizer · CMF rewrite (76.0%) ? Rephrasy rewrite (43.3%) - Application essay ? S26
Humanizer · CMF rewrite (80.0%) ? Rephrasy rewrite (62.7%)
Scope and source
The commercial entry uses historical published rewrites, and the open-weight entry uses outputs collected from an identified model artifact. The comparison uses the recorded outputs in the November 2026 dataset, version 2026-11-v1. See the methodology for metric definitions and limitations.
Model profiles: Humanizer · CMF and Rephrasy.