October 2025 (microsoft/onnxruntime-genai): Delivered TRT-RTX execution provider support with a user-facing NvTensorRtRtx alias and enforced default QDQ quantization when TRT-RTX is selected. This improves quantization path consistency, stabilizes model-building behavior, and aligns CLI argument parsing with TRT-RTX usage. The work enhances reliability and integration for GenAI workloads and sets the foundation for further TRT-RTX optimizations.

Monthly summary for 2025-09: Implemented user-facing name alignment for NvTensorRTRTXExecutionProvider in microsoft/Olive, renaming to NvTensorRtRtx to reflect GenAI naming discussions and ensure consistent branding in the ModelBuilder API. This refactor aligns internal representation with external naming, reducing user confusion and improving clarity for downstream integrations across Olive workflows.

November 2024 highlights for microsoft/Olive: Delivered refined AWQ INT4 quantization integration and enhanced TensorRT Model Optimizer workflow, improving low-precision inference readiness and deployment stability. Removed outdated BERT example, added Phi-3 model example, and tightened opset version handling. Updated dependencies and documentation to streamline Windows RTX deployments. Business impact: reduced setup friction, faster model optimization cycles, and clearer guidance for quantized inference on Windows hardware.