TheFinAI/Fin-o1-14B

Hugging Face
TEXT GENERATIONPricing:Input $0.48 / Output $0.96Concurrent Unit Cost:1Model Size:14BQuant:FP8Context Size:32kTool Calling:SupportedPublished:May 15, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Gated Featherless Exclusive Warm

TheFinAI/Fin-o1-14B is a 14 billion parameter financial reasoning model developed by TheFinAI, fine-tuned from Qwen3-14B. It specializes in financial numerical reasoning and question answering over financial text and tables, utilizing supervised fine-tuning on financial chain-of-thought data and GRPO reinforcement learning. With a context length of 40,960 tokens, this model is designed for research in financial AI and serves as a strong baseline for financial reasoning tasks.

Loading preview...

Overview

Fin-o1-14B is a 14 billion parameter financial reasoning model developed by TheFinAI, built upon the Qwen3-14B base model. It is specifically fine-tuned for financial applications, leveraging supervised fine-tuning (SFT) on financial chain-of-thought data and further enhanced with GRPO reinforcement learning. This model is part of the newer Fin-o1 generation, which is recommended over the earlier Fino1 series for new work.

Key Capabilities

  • Financial Numerical Reasoning: Excels at complex calculations and quantitative analysis within financial contexts.
  • Question Answering: Capable of answering questions based on financial text and tabular data.
  • Research Baseline: Serves as a robust foundation for further research and development in financial AI.
  • High Context Length: Supports a substantial context window of 40,960 tokens, allowing for processing of longer financial documents.

Good For

  • Academic Research: Ideal for researchers exploring financial numerical reasoning and the application of LLMs in finance.
  • Benchmarking: Useful as a strong baseline model for evaluating new financial reasoning models.
  • Financial Data Analysis: Can assist in tasks requiring interpretation and calculation from financial reports and statements.

Limitations

  • Not for Investment Advice: The model is not intended for making investment decisions or providing financial advice without human oversight.
  • Potential for Errors: May produce incorrect calculations or hallucinated figures, particularly with lengthy documents or multiple tables.
  • English Only: Currently trained and evaluated exclusively on English data.

Finetunes
2 models