Fino1-4B: Financial Reasoning Language Model

Fino1-4B is a specialized 4 billion parameter language model developed by khazarai, fine-tuned from the Qwen3-4B architecture. Its primary focus is on financial reasoning and question answering in English.

Key Capabilities

• Structured Financial Reasoning: Provides step-by-step, explainable answers to financial questions.
• Data-driven Insights: Adapted using LoRA fine-tuning on a dataset derived from FinQA, enhanced with GPT-4o-generated reasoning paths.
• Domain-Specific Expertise: Optimized for the finance domain, making it suitable for analyzing financial documents and data.

Good For

• Financial Report Analysis: Interpreting and extracting information from financial reports.
• Numerical Reasoning: Performing calculations and logical deductions over tabular financial data.
• Educational & Research Q&A: Generating structured answers for financial inquiries.

Limitations

• Not Financial Advice: Outputs should not be considered professional financial advice.
• Potential for Hallucinations: May misinterpret complex financial contexts or hallucinate numbers.
• Coverage: Trained on approximately 5.5K examples, which may limit its coverage for highly niche financial instruments or international standards.