Fabrix-AI-Inc/Argos-2B

VISIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2.3BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 27, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Argos 2B by Fabrix-AI-Inc is a 2.3 billion parameter Qwen3_5ForConditionalGeneration model, specifically fine-tuned for knowledge retrieval and documentation citation within the Fabrix.ai / RDAF ecosystem. It utilizes Supervised Fine-Tuning (SFT) on a proprietary knowledge base and Direct Preference Optimization (DPO) to ensure accurate card citations and precise custom widget dashboard pack generation. This model excels at generating repo-relative knowledge base paths, guaranteeing 100% card format accuracy and high custom widget pack order accuracy.

Loading preview...

Argos 2B: Specialized Knowledge Retrieval and Citation Model

Argos 2B, developed by Fabrix-AI-Inc, is a 2.3 billion parameter model built on the Qwen3_5ForConditionalGeneration architecture. It is uniquely adapted for specialized knowledge retrieval and documentation citation tasks within the Fabrix.ai / RDAF environment, focusing on generating precise references and custom widget dashboard configurations.

Key Capabilities and Alignment

  • Domain-Adapted Fine-Tuning: The model undergoes Supervised Fine-Tuning (SFT) using the extensive Fabrix / RDAF knowledge base and custom widget recipes.
  • Direct Preference Optimization (DPO): DPO is applied to enforce strict card citations (kb/cards/*.md) and ensure exact 34-item ordered card dependencies for custom widget dashboards.
  • High Citation Accuracy: Achieves 100% card format accuracy and 81.2% custom widget pack order accuracy, significantly outperforming baseline SFT models.
  • Context Length: Supports a context length of 32768 tokens.

Ideal Use Cases

  • Knowledge Retrieval: Generating accurate, repo-relative kb/ paths for documentation and knowledge base articles.
  • Documentation Citation: Ensuring precise citation of kb/cards/*.md files based on user queries.
  • Custom Widget Dashboard Generation: Creating ordered lists of card dependencies for custom widget dashboards with high fidelity.
  • Automated Response Generation: Providing concise, citation-focused responses without prose or extraneous information, as demonstrated in the quickstart example for system prompts like "Fabrix cite-mode retrieval."

Finetunes
2 models