Synaxi-predict: I'm trying to predict token cost before it happens (github.com)

Synaxi has launched "synaxi-predict," a tool that forecasts the cost, turn count, and success rate of tasks run through the Claude AI model, allowing users to select the most suitable model before execution. This feature significantly enhances operational efficiency by reducing token waste—an essential concern given the rising costs associated with AI model usage. By combining this predictive capability with Synaxi's existing macOS application, which improves Claude API cost management by streamlining request formats and eliminating token redundancy, users can now minimize unnecessary expenses both from model selection and execution. The tool employs a multi-layer perceptron (MLP) trained on approximately 53,000 previous runs to generate accurate predictions. It incorporates key input features such as task descriptions and code complexity metrics sourced through tree-sitter, aiding in the prediction process. Each completed task updates the model with actual results, refining its accuracy over time. This feedback loop fosters better model selection and cost efficiency in real-world applications, making "synaxi-predict" a significant advancement for developers using AI in coding tasks. With its automated integration into coding sessions and valuable community contributions for retraining, "synaxi-predict" stands to enhance the overall AI/ML ecosystem by promoting smarter, cost-effective use of AI resources.