tjarvis91/vfaix-vpa-options-trader

VFAi-X Franken-B: Vision-Language Options Trading Model

This model, Franken-B, is a 9 billion parameter vision-language model built upon Qwen3.5-9B-VL by tjarvis91. It is specifically designed for US equity options trading, processing composite chart images (1D/1H/5M) alongside structured market data to output trading decisions (BUY/SELL/HOLD/NO_TRADE) with conviction and a risk plan. The model is optimized for Volume-Price Analysis (VPA), pattern recognition, and chart-text fusion, demonstrating significant improvements over its predecessor, V3.7.

Key Capabilities

• Multimodal Analysis: Integrates visual chart data with textual market context for comprehensive trading signals.
• Options Trading Focus: Generates specific trading actions and risk plans tailored for options markets.
• Performance Gains: Achieved +6,813 pp 2-year return and +106.76% vision-only fusion compared to V3.7 in simulated backtests.
• Composition-Lock Recipe: Utilizes a unique stacking of three LoRA adapters (V3.7s base → V5.0 LM LoRA → V5.8 last-3 LM LoRA) on Qwen3.5-9B-VL, where adapter order is critical for performance.
• FP8 Optimization: Designed for efficient inference using FP8 quantization on vLLM, suitable for consumer GPUs.

Good For

• Researchers studying multimodal AI for financial markets.
• Developers building algorithmic trading systems requiring vision-language capabilities.
• Exploring advanced chart pattern recognition and Volume-Price Analysis in an AI context.

Note: This model represents a historical lineage, superseded by the Qovaryx Options Decoder app for current trading inference. It is retained for reproducibility and research into its unique composition-lock methodology.