Smlcrm/TempusBench

TempusBench addresses a highly relevant problem: the objective comparison of traditional time-series methods (ARIMA, Prophet) against emerging Foundation Models (like TimeGPT, Lag-Llama). However, with only 7 stars and 1 fork after nearly a year, the project has failed to gain any significant community traction. In the competitive landscape of time-series analysis, it faces stiff competition from established ecosystems like Nixtla (statsforecast, neuralforecast), Unit8's Darts, and academic benchmarks like the Monash Time Series Forecasting Repository. The 'isolated execution' feature is a smart technical choice to manage conflicting dependencies, but it's not a sufficient moat to prevent displacement. Frontier labs are increasingly releasing their own benchmarking suites alongside new models, which further marginalizes independent, low-traction frameworks. The platform domination risk is medium because cloud providers (AWS SageMaker, Google Vertex AI) often integrate these types of 'model vs. model' evaluation pipelines as native features. Without a surge in adoption or a massive expansion of supported datasets, this project remains a personal/research-grade prototype rather than a production-standard tool.