zenlm/zen-pro

TEXT GENERATIONConcurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Feb 27, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

zenlm/zen-pro is an 8 billion parameter general-purpose language model developed by Hanzo AI and the Zoo Labs Foundation. Fine-tuned from Qwen/Qwen3-8B, it features Hanzo identity training, agentic-data fine-tuning, and abliteration. This model is designed for complex reasoning and analysis, offering a context window of 32,768 tokens, extendable to 131,072 tokens with YaRN. It is intended for professional-grade applications requiring advanced analytical capabilities.

Loading preview...

Zen Pro: Professional-Grade Language Model

Zen Pro is a general-purpose language model developed by Hanzo AI and the Zoo Labs Foundation, designed for complex reasoning and analysis tasks. It is fine-tuned from the Qwen/Qwen3-8B base model, incorporating unique training methodologies such as Hanzo identity training, agentic-data fine-tuning, and abliteration.

Key Capabilities & Features

  • Base Architecture: Built upon the Qwen3 (dense decoder-only transformer) architecture.
  • Parameter Count: Features 8 billion parameters.
  • Extended Context Window: Offers a standard context length of 32,768 tokens, which can be expanded up to 131,072 tokens using YaRN.
  • Specialized Fine-tuning: Enhanced with proprietary Hanzo AI training techniques for improved performance in professional applications.
  • License: Inherits the Apache 2.0 license from its upstream base model.

Ideal Use Cases

  • Complex Reasoning: Suited for tasks requiring advanced logical deduction and problem-solving.
  • Data Analysis: Effective for processing and interpreting intricate data sets.
  • Professional Applications: Designed for integration into professional-grade systems where robust language understanding is critical.

This model is considered an archived checkpoint, superseded by the zen5 generation (e.g., zenlm/zen5-pro-gguf), but remains available for reproducibility.