Supported operating system:

Windows 11 64-bit

Release note:
- Enabled and optimized Granite 4.0-1B model support on NPU.
- Improved NPU performance for Qwen3-4B model workloads.
- Added support and performance optimizations for Granite 4.0-1B on NPU.
- Enhanced NPU performance for Qwen3-TTS-0.6B model workloads.
- Added support for Gemma Embedding model workloads.
- Fixed a compiler issue that could cause incorrect results with grouped INT8 quantized weights.
- Optimized NPU performance for Qwen Omni INT4 TTS workloads.
- Enabled and optimized Qwen3-ASR-0.6B model support on NPU.
- Improved NPU performance for MinerU-Pro model workloads.
- Enhanced overall model accuracy and reliability for supported AI workloads.
- Resolved an access violation issue identified through telemetry monitoring.
- Fixed an NPU memory management issue that could cause crashes during Phi model inference.
- Resolved a telemetry-reported invalid pointer access issue affecting NPU stability.

Supported Devices:
Intel(R) NPU
Intel(R) Reserved Device