AMD Releases ROCm 10 With AI Developer Tools
AMD released ROCm 10 on August 27, making its ROCm.AI developer experience generally available alongside a broader software-stack update. AMD says the release adds a modular Core SDK and tools including Hyperloom, Skills and ROCm CLI; its performance comparisons are vendor results tied to a specified test configuration.
AMD released ROCm 10 on August 27, presenting the update as a broader refresh of its GPU software stack and making the ROCm.AI developer experience generally available. The release combines a more modular ROCm Core SDK with updated libraries, compilers, framework support and developer tooling.
Developer tooling and workflow changes
ROCm.AI brings together three named components: AMD Skills, ROCm CLI and ROCm Hyperloom. AMD describes Hyperloom as an agentic system that profiles an inference workload, identifies bottlenecks, explores optimizations, implements changes and validates the result. ROCm CLI remains a technology-preview interface for setting up and operating AI workloads.
AMD also says its Skills packages are available through Claude Code, Codex and Cursor marketplaces as well as an open GitHub catalog. The company positions the components as ways to reduce manual setup, diagnostics and optimization work for teams using AMD hardware.
Platform updates and performance claims
The ROCm 10 release notes list updates across inference, developer tooling and profiling, plus framework support that includes PyTorch, JAX, vLLM, SGLang, TensorFlow, MIGraphX and ONNX Runtime. AMD's launch material reports average 3.3x inference and 2.4x training improvements over ROCm 7 on the same configuration; those are vendor measurements and should be evaluated against each team's workload and hardware.
Compatibility and support details should likewise be checked against the release documentation, because supported hardware and framework versions vary by platform.
The release is relevant to organizations already running, or considering, AMD Instinct infrastructure because software compatibility and operational tooling can shape deployment time. It does not by itself establish performance for every model, framework or GPU configuration.
Key Points
- 1ROCm 10 makes AMD's ROCm.AI developer experience generally available and updates the wider GPU software stack.
- 2The release introduces or expands AMD Skills, ROCm CLI and ROCm Hyperloom for AI-workload setup and optimization.
- 3AMD's published performance comparisons are vendor measurements tied to a specific configuration.
Scoring Rationale
ROCm 10 is a notable platform release for teams considering or operating AMD GPU infrastructure, particularly because it combines serving packages, framework distribution, diagnostics, and assistant-oriented workflows. Its reported performance improvements are consequential but are vendor benchmark results tied to a specific MI355X configuration and model set.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
