{"id":"smaug-humanizer","slug":"smaug-humanizer","name":"Smaug Humanizer","orgId":"abacus-ai","family":"smaug-humanizer","releaseDate":null,"availability":"announced","openWeights":true,"license":null,"paramsTotal":null,"paramsSource":null,"contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://blog.abacus.ai/blog/2026/10/26/smaug-humanizer-an-open-weight-12b-humanizer/","hfRepo":null,"modalities":["text"],"aliases":["Smaug Humanizer","smaug humanizer","smaug-humanizer","abacusai/Smaug-Humanizer","abacusai/Smaug-Humanizer-GGUF"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-10-27","announced":"2026-10-26","blurb":"Abacus.AI's open-weight 12B model that rewrites AI-generated text so it reads as if a person wrote it, keeping the facts, numbers, names and layout while changing wording, rhythm and voice. Fine-tuned from Google's Gemma 4 12B and released under Apache 2.0 on Hugging Face, with GGUF builds.","api":"none","openWeights":"yes","params":{"text":"12B (11,959,730,224 safetensors parameters)","source":"Hugging Face model card and metadata (abacusai/Smaug-Humanizer)"},"facts":[{"label":"Type","value":"Text rewriting model: rewrites AI-generated text into a human style"},{"label":"Base model","value":"google/gemma-4-12B (the pretrained model, not instruction-tuned)"},{"label":"Training","value":"SFT with LoRA (r=64, alpha=128, 2 epochs) on 3,676 AI-to-human rewrite pairs, then 200 GRPO steps with a kernel-similarity (MMD) style reward and five deterministic gates (numbers, layout, length, script, spelling)"},{"label":"Weights","value":"Hugging Face: abacusai/Smaug-Humanizer (BF16) and abacusai/Smaug-Humanizer-GGUF (Q8_0 fits a 16 GB GPU; Q4_K_M runs on an 8 GB GPU)"},{"label":"License","value":"Apache 2.0"},{"label":"Usage","value":"Split long inputs into chunks of about 500 words; sampling temperature 1.0, top_p 0.95, top_k 64, stop on <end_of_turn>"},{"label":"Languages","value":"Mostly English; other languages work less well, per the model card"},{"label":"API","value":"None announced; self-hosted (vLLM, llama.cpp, Ollama, LM Studio)"}],"claims":["Lab claim: turns many texts that the Pangram AI-text detector scores as AI into text it scores as human, though not every time; long texts are flagged more often."],"sources":[{"kind":"lab","title":"Smaug-Humanizer: An open-weight 12B humanizer (Abacus.AI blog)","url":"https://blog.abacus.ai/blog/2026/10/26/smaug-humanizer-an-open-weight-12b-humanizer/","date":"2026-10-26"},{"kind":"lab","title":"Bindu Reddy (Abacus.AI CEO) announcing Smaug Humanizer on X","url":"https://x.com/bindureddy/status/2114938847704571975","date":"2026-10-27"},{"kind":"repo","title":"abacusai/Smaug-Humanizer on Hugging Face","url":"https://huggingface.co/abacusai/Smaug-Humanizer","date":"2026-10-26"},{"kind":"repo","title":"abacusai/Smaug-Humanizer-GGUF on Hugging Face","url":"https://huggingface.co/abacusai/Smaug-Humanizer-GGUF","date":"2026-10-26"}],"firstBenchmark":null},"organization":{"id":"abacus-ai","name":"Abacus.AI","slug":"abacus-ai","country":null,"url":null,"color":"#11110F","mark":"A"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Oct 26, 2026; no independent results yet.","lead":"As of Nov 3, 2026, Smaug Humanizer is pending benchmarks on isitsota: announced Oct 26, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}