FinGPT

FinGPT is an open source financial language model project from the AI4Finance Foundation. Rather than training a finance model from scratch, it adapts existing open models to financial text and ships the weights, training scripts, datasets and benchmarks alongside managed APIs and deployment support. The aim is to let research groups and engineering teams build financial AI without the budget a ground up model would require. The platform covers sentiment analysis over news, filings and market narratives; forecasting signals, including a public demo that takes a ticker, a start date and one to four weeks of history and returns a predicted price movement with news analysis and a financial summary; retrieval workflows grounded in filings and structured data; multi agent analysis spanning sentiment, market data and fundamentals; extraction of metrics and risks from earnings and annual reports; and benchmarking across finance specific datasets and tasks. Models are distributed through GitHub, HuggingFace and ModelScope, and enterprise deployment covers on premise installation packages, private cloud images and configuration guides. Three plans are published. Free costs $0 and includes the open source models and code, community support through GitHub and Discord, benchmark dataset access, a sentiment API capped at 100 requests per day and a forecaster API capped at 10 per day. Premium is $19.9 per month and raises those caps to 10,000 and 1,000 requests per day, adding custom fine tuning support, priority email support and a 99.5% uptime service level agreement. Enterprise is custom priced through sales and covers dedicated deployment, custom models and organizational controls. Cost control is the project's stated argument. Its published comparison puts instruction tuning one of its models at roughly $65.6 of GPU time, against figures in the millions for training comparable financial models from scratch.