November 2026 · Independent humanizer benchmarks. Explore the evidence

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.

Pick 2 to 4 entries 1 selected

November 2026
Smaug HumanizerVerified 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.abacusai/Smaug-Humanizer
Open weights
Shared score components
Overall score44.18
Panel bypass score18.5%
Meaning preservation76.5%
Raw meaning cosine0.9414
Readability81.3%
Category consistency88.2%
Detector coverage231/231
Identical outputs0/33
Penalty points−1
Per-detector human-likelihood scores · measured outputs only
GPTZero4.8% 33/33
ZeroGPT0.7% 33/33
Copyleaks27.3% 33/33
Originality.ai5.5% 33/33
Pangram25.6% 33/33
Grammarly50.7% 33/33
QuillBot99.5% 33/33
Category scores · before entry-level penalties
Academic essay44.3
Application essay41.0
Blog post43.0
Business email43.3
Marketing copy58.0
Discussion board43.0
News article38.1
EvidenceAll 33 outputs