Back in July ROCm 7.14 was announced as their new production release built atop TheRock build system and introducing Ryzen AI 400 series support. The versioning choice was a bit odd since ROCm 7.2 was the stable release at the time while ROCm 7.9+ turned into "tech preview" releases. With ROCm 7.14, it was declared stable. In clearing up that versioning mess, they have done a major version bump... to ROCm 10.0.

ROCm 10.0 is succeeding the ROCm 7.x series as the newest major release of this open-source AMD Radeon/Instinct GPU compute stack. Most significant with ROCm 10.0 is the new ROCm.AI for AI-driven development.

Plus the introduction of the ROCm CLI.

With this big version bump to ROCm 10.0, this also marks their first release of ROCm since shifting to a six week release cycle moving forward. Now will the next release in six weeks be ROCm 10.0 or ROCm 11.0 given today's versioning trends? We'll see.

ROCm 10.0 also has Hyperloom as announced at last month's AMD Advancing AI event. There is also improved consolidation around their Linux packages for ROCm, production-ready containers and Python wheels for vLLM, local LLM fine-tuning for Ryzen AI Max, and other improvements.

AMD on Windows also retired their HIP SDK package in favor of the ROCm Core SDK. Windows will follow the same release cadence as Linux for ROCm releases. Right now the ROCm Core SDK for Windows is a simple static package while later in the year they are working toward native Windows installer support.

More details on ROCm 10.0 can be found via today's announcement on rocm.blogs.amd.com.