tjarvis91/vfai-x-3.5-9b-options

VISIONConcurrency Cost:1Model Size:9BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:May 12, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

The tjarvis91/vfai-x-3.5-9b-options model is a 9 billion parameter vision-language model, fine-tuned from Qwen/Qwen3.5-9B-VL, specifically designed for quantitative finance and options trading analysis. It excels at chart pattern recognition and technical analysis, serving as a core component for automated trading decision-making. This model is optimized for local-first, sub-millisecond CPU inference, making it suitable for integration into desktop applications for real-time financial market analysis.

Loading preview...

VFAi-X 3.5 9B Options: Specialized Financial Vision-Language Model

This model, developed by tjarvis91, is a 9 billion parameter vision-language model (VLM) fine-tuned from Qwen/Qwen3.5-9B-VL, specifically engineered for quantitative finance and options trading. It is a legacy lineage model, with its core functionality now integrated into the Qovaryx Options Decoder desktop application, which offers sub-millisecond CPU inference without requiring a GPU.

Key Capabilities & Features

  • Multimodal Financial Analysis: Processes both textual and visual data (e.g., stock charts) for comprehensive trading insights.
  • Specialized Trading Models: Includes two distinct branches for research and ensemble building:
    • V3.7 (main branch): Excels in text-heavy reasoning and small-N corpus discipline, demonstrating 100% brutal action_pass and strong performance on broad small-N corpus coverage. It is particularly effective for text-only fusion tasks.
    • V6 Sniper (vfai-x-sniper-options branch): A high-conviction abstention specialist that, despite a low trade take-rate (~2%), yields the highest per-trade profit factors in the VFAi-X family, especially for penny stocks (37.18 PF). It serves as an excellent confirmation filter.
  • Local-First Operation: Designed for desktop application integration, supporting local inference with vLLM acceleration and FP8 quantization.
  • Broker Integration: Features a Tradier broker bridge for user-supplied paper/live credentials within the associated desktop app.

Good For

  • Quantitative Finance Researchers: Ideal for studying specialized trading model lineages, ensemble building, and exploring high-conviction trading strategies.
  • Algorithmic Trading Development: Provides components for automated options trading systems, particularly for chart analysis and decision support.
  • Developers of Financial Applications: Suitable for integration into local-first desktop applications requiring fast, specialized financial inference capabilities.