Infinix-io/humanizer

TEXT GENERATIONPricing:Input $1.2 / Cached $0.24 / Output $4.8Concurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 17, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Infinix-io/humanizer is a 12 billion parameter model designed to rewrite AI-generated text into more human-like prose, supporting both English and Chinese. Developed by Infinix-io, it maintains factual accuracy by preserving numbers, units, dates, names, and quotes while rephrasing content to avoid sounding artificial. This model is optimized for local execution and excels at humanizing various drafts like emails, essays, and reports, making it ideal for users seeking to refine AI output for natural readability.

Loading preview...

humanizer: AI-to-Human Text Rewriter

humanizer is a 12 billion parameter model developed by Infinix-io that specializes in transforming AI-generated drafts into text that reads as if a human wrote it. It supports both English and Chinese, making it versatile for a global user base. A key differentiator is its ability to maintain strict factual fidelity, ensuring that all numbers, units, dates, names, and quotations remain unchanged during the rewriting process. The model is designed for local execution, with optimized GGUF files available for various memory configurations, allowing it to run efficiently on machines with as little as 8 GB of RAM.

Key Capabilities

  • Human-like Rewriting: Rewrites AI-generated text (emails, essays, reports, forum posts) to sound natural and human, varying sentence length and cutting hedging.
  • Factual Preservation: Strictly preserves all facts, numbers, units, dates, names, and quotations, adding no new information.
  • Multilingual Support: Processes drafts in both English and Chinese.
  • Local Execution: Optimized for offline use with dedicated applications for Mac and Windows, and compatible with llama.cpp, MLX, and transformers runtimes.
  • High AI-Detector Evasion: Achieves 95% human judgment by Originality.ai at its strictest setting, significantly outperforming previous versions and public baselines.
  • Quantization-Aware Training: Features custom-quantized GGUF files (2-bit, Q3, Q4_K_M) that offer improved performance over standard llama.cpp quantization at smaller file sizes.

Good for

  • Content Creators: Refining AI-generated articles, blog posts, or marketing copy to enhance readability and authenticity.
  • Professionals: Humanizing AI-drafted emails, reports, and presentations for a more personal and engaging tone.
  • Students: Polishing AI-assisted essays and assignments to meet academic standards for natural language.
  • Developers: Integrating into applications or agents that require human-quality text output from AI models, with clear prompt formatting and sampling parameters provided.

Finetunes
1 models