FinGPT — Open-Source Financial Large Language Model

FinGPT — Open-Source Financial Large Language Model

FinGPT is an open-source financial language model framework from the AI4Finance Foundation. Rather than training from scratch, it leverages instruction fine-tuning on top of open base models (LLaMA, Falcon, MPT) using a curated corpus of financial text — news, 8-Ks, earnings call transcripts, analyst reports, and Reddit/StockTwits sentiment — enabling tasks like sentiment analysis, event extraction, credit scoring, and return forecasting at a fraction of proprietary LLM cost.

Key Features

• Task-specific LoRA adapters: financial sentiment, headline classification, NER, return prediction
• Live financial data pipeline (news, Reddit, SEC EDGAR, Yahoo Finance)
• RLHF pipeline for aligning models to financial expert feedback
• Model zoo on HuggingFace (fingpt-sentiment, fingpt-forecaster, fingpt-mt, fingpt-analyst)
• Online deployment recipes for vLLM and HuggingFace Inference Endpoints

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf")
model = PeftModel.from_pretrained(base, "FinGPT/fingpt-sentiment_llama2-7b_lora")
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")

prompt = "Analyze the sentiment: Apple beats Q3 earnings expectations."
inputs = tokenizer(prompt, return_tensors="pt")
print(tokenizer.decode(model.generate(**inputs, max_new_tokens=64)[0]))

npx ai-supply add fingpt-financial-llm

Curated mirror of the open-source FinGPT (MIT). Get it from the source.