This study investigates the integration of real-time sentiment analysis from financial news using large language models (GPT-2 and FinBERT) with traditional technical indicators (MACD, SAR, VW MACD, Dual MACD) and time-series forecasting models (ARIMA, Prophet, ETS) to enhance S&P 500 trading strategies. The authors collect daily financial news from major outlets and S&P 500 price data from 2019 to 2024, aligning news sentiment with next-day returns using a one-day lag to respect market close. Sentiment scores are derived from fine-tuned GPT-2 and FinBERT, while technical indicators capture momentum and trend patterns. Time-series models forecast price movements based on historical data. A comprehensive set of trading strategies is evaluated, including a benchmark buy-and-hold, sentiment-only, technical-only, and hybrid approaches that combine all components. Results demonstrate that hybrid strategies significantly outperform benchmarks, achieving higher cumulative returns and better adaptability to volatile market conditions. The study highlights the complementary nature of sentiment and technical analysis, with FinBERT providing more precise financial sentiment due to its domain-specific pre-training. The findings suggest that integrating diverse data sources and models can lead to more robust and profitable trading systems, offering practical insights for quantitative traders and financial analysts.