BotBeat
...
← Back

> ▌

AMDAMD
UPDATEAMD2026-05-23

AMD Lemonade SDK 10.5 Elevates macOS Support to General Availability with ROCm 7.13

Key Takeaways

  • ▸macOS support elevated to GA status with all major Lemonade SDK features now production-ready for Apple Silicon systems
  • ▸ROCm 7.13 integration brings support for AMD Instinct MI350P accelerators and Ubuntu 26.04 LTS
  • ▸Enhanced model management and upstream library updates improve developer experience for optimizing LLMs on various hardware
Source:
Hacker Newshttps://www.phoronix.com/news/Lemonade-SDK-10.5-Released↗

Summary

AMD has released Lemonade SDK 10.5, promoting macOS support from beta to general availability (GA) status. The milestone reflects improved quality and stability for Apple Silicon-based systems, with all major features now considered production-ready for macOS users despite AMD hardware not being present in modern Apple Macs. Support is maintained through the Llama.cpp Metal backend for Apple's M-Series GPUs.

The release integrates ROCm 7.13, AMD's open-source GPU computing platform, which adds support for the new AMD Instinct MI350P accelerator, Ubuntu 26.04 LTS compatibility, and broader support for existing AMD Ryzen AI and Radeon PRO hardware. Version 10.5 also delivers improved custom model management, updates to Llama.cpp (9174) and Stable-Diffusion.cpp libraries, and various bug fixes.

Lemonade SDK continues its momentum since version 10.0 launched in March 2026, which enabled AMD Ryzen AI NPUs to effectively run large language models on Linux. The open-source AI framework, designed for 'refreshingly fast local AI,' is seeing increasingly frequent releases alongside complementary tools like AMD's GAIA software.

  • Lemonade SDK maintains rapid release cadence with increasingly frequent updates since reaching v10.0 in March 2026
Large Language Models (LLMs)Generative AIMLOps & InfrastructureOpen Source

More from AMD

AMDAMD
PRODUCT LAUNCH

AMD's Ryzen AI Halo Makes Local AI Development Accessible, But at a Premium Price

2026-07-06
AMDAMD
RESEARCH

AMD MI355X Proves Competitive for Frontier AI Inference at 2.75x Lower Cost Than Blackwell

2026-07-03
AMDAMD
RESEARCH

Stanford Researchers Develop Multi-Agent AI System to Improve HIP Kernel Generation for AMD GPUs

2026-07-02

Comments

Suggested

AnthropicAnthropic
RESEARCH

Claude Code's Trust Model Exposes Developers to Malicious Hooks, Security Analysis Warns

2026-07-07
DeepSeekDeepSeek
RESEARCH

Verification Loops Give DeepSeek 4x Boost, Matching Opus at 1/7 the Cost

2026-07-07
Corvin LabsCorvin Labs
PRODUCT LAUNCH

CorvinOS Launches Self-Hosted Agentic OS With EU AI Act Compliance Baked Into Architecture

2026-07-07
← Back to news
© 2026 BotBeat
AboutPrivacy PolicyTerms of ServiceContact Us