Developer Resource
Accelerate your AI development with curated playbooks and a connected developer community.
Go From Idea to Workflow in Minutes
Start building faster, from power-on to tokens in minutes, with pre-configured software, regular updates, OS flexibility, and expert support.
Set up anywhere and run intensive AI workloads locally with AMD Ryzen AI Halo, featuring 128 GB unified memory and support for up to 200B parameter models in a compact, power-efficient design, without constraints of cloud costs.
The Ultimate AI Developer Platform
AMD ROCm™ software powers class-leading Generative AI performance and Day 0 support.
Apple M4 Pro vs. AMD Ryzen™ AI Halo1
Apple M4 Pro
AMD Ryzen AI Max
NVIDIA DGX Spark vs. AMD Ryzen™ AI Halo2,3
GLM 4.7 Flash-30B-A3B
GPT-OSS
-120B
Qwen 3.5-122B-A10B
Qwen 3.6-35B-A3B
NVIDIA DGX Spark
AMD Ryzen AI Halo
Performance (Tokens/Second)
NVIDIA DGX Spark
Supports Linux Only
Lower Tok/Sec per $
No NPU
AMD Ryzen AI Halo
Supports Windows and Linux OS
Leadership LLM Tok/Sec per $
Up to 50 TOPS NPU4
Build, Run, and Deploy Your Way
Real Workflows, Ready to Run
Instant productivity with streamlined onboarding experience, clear guides and ready-to-run apps for image generation, coding assistants, automations, LLMs and more.
Run Bigger AI Locally
Experiment, prototype, and deploy faster – without constraints of cloud costs. With both Linux and Windows support, developers can power their entire journey from experiment to production.
Build AI at Any Stage
From first run to fine-tuning, AMD Ryzen AI Halo supports AI for everyone. Explore hands-on training from beginner to advanced AI workflows and receive prioritized AMD support in GitHub and the AMD AI Developer Program.
AI That Works For You
Unlock real‑time reasoning, persistent memory, and seamless automation across apps and browsers. No token tax.
Pay Once For Cloud Grade Intelligence4
Run sustained AI inference locally with a one-time hardware investment and low estimated monthly electricity cost.
Up to
6x
lower 3-year cost
vs. equivalent cloud API usage in a modeled sustained-use scenario.
Enhance the Developer Experience with Software from AMD
AMD Ryzen AI Halo is built for open frameworks and comes pre-installed with the AMD Ryzen AI Halo Developer Center app, a central hub for developers to easily access playbooks, tools and updates.
Image Zoom
Getting started has never been easier. Take advantage of AMD AI Playbooks, guided learning path for developers, including quick and easy setups.
Image Zoom
Coming soon, AMD Sync is a powerful desktop application that lets you seamlessly monitor and manage your AMD Ryzen AI Halo from another device. Access and launch development tools and effortlessly sync files between your machines.
Image Zoom
Discover curated AMD resources, stay aligned with validated software configurations for optimal performance, and seamlessly access and manage pre-installed tools, all from the AMD Ryzen AI Developer Center.
What's Under the Hood
AMD Ryzen AI Halo Developer Platform with Ryzen AI Max+ 395 processor with full ROCm software support.
128GB
LPDDR5x Unified Memory
60 FP16 TFLOPS
GPU Performance
Windows or Linux
Operating System
Up to
50 TOPS
NPU Performance⁵
Coming Soon: AMD Ryzen AI Halo equipped with the AMD Ryzen™ AI Max+ PRO 495 processor with 192GB memory support.
Resources
This product is available for purchase and use in the United States. It has been designed and tested to comply with applicable U.S. regulatory requirements.
Footnotes
© 2026 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, AMD ROCm, Radeon, Ryzen, and combinations thereof are trademarks of Advanced Micro Devices, Inc. Linux® is the registered trademark of Linus Torvalds in the U.S. and other countries. Windows is a registered trademark of Microsoft Corporation in the US and/or other countries. Other product names used in this publication are for identification purposes only and may be trademarks of their respective owners. Certain AMD technologies may require third-party enablement or activation. Supported features may vary by operating system. Please confirm with the system manufacturer for specific features. No technology or product can be completely secure.
- SHO-50: Testing as of May 2026 by AMD. Performance based on multiple runs of Pytorch-based workloads using default templates in ComfyUI 0.8.36. Workloads tested: Stable Diffusion XL, Flux Schnell, Flux 2. Klein, Qwen Image, Qwen Image Edit, Z Image Turbo, Ace Step 1.5, Ace Step 1.5 XL, Hunyuan 3D 2.1, LTX 2.3, and Wan 2.2. System specifications: Pre-production AMD Ryzen™ AI Halo developer platform with AMD Software: Adrenalin Edition™ 26.10.08 drivers vs. Apple Mac Mini with Apple M4 Pro processor, 12-core CPU, and 16-core GPU, MacOS Tahoe 26.4, 64GB memory and 512GB storage set to High Performance Mode. Performance may vary. SHO-5
- SHO-57: Testing as of May 2026 by AMD. Performance based on average across 3 runs of GPT OSS 120B model, Qwen 3.5 122B Model, Qwen 3.6B Model and GLM 4.7 Flash 30B Model. Testing completed on pre-production AMD Ryzen AI Halo developer platform with AMD Ryzen AI Max+ 395 processor, 128GB LPDDR5x memory, and Linux OS. Performance compared against NVIDIA DGX Spark, with 128GB LPDDR5x using latest software stack available as of May 6th 2026.Tokens per second calculated based on 100 token context. System manufacturers may vary configurations, yielding different results. SHO-57
- SHO-61: Testing as of May 2026. Retail price for NVIDIA DGX Spark is $4699, Retail price for AMD Ryzen AI Halo is $3999. Performance based on average across 3 runs of GPT OSS 120B model, Qwen 3.5 122B Model, Qwen 3.6B Model and GLM 4.7 Flash 30B Model. Testing completed on pre-production AMD Ryzen AI Halo developer platform with AMD Ryzen AI Max+ 395 processor, 128GB LPDDR5x memory, and Linux OS. Performance compared against NVIDIA DGX Spark, with 128GB LPDDR5x using latest software stack available as of May 6th 2026.Tokens per second calculated based on 100 token context. System manufacturers may vary configurations, yielding different results. SHO-61
- SHO-49: Specimen use-case using simplified assumptions. Cloud comparison assumes Claude Sonnet 4.5 standard API pricing of $3/M input tokens and $15/M output tokens, with a 10:1 input-to-output token utilization mix. Local scenario assumes AMD Ryzen™ AI Max+ sustained throughput at 128K context of 36 output tokens/sec and 446 input tokens/sec, measured on pre-production AMD Ryzen™ AI Halo developer platform using LM Studio 0.4.12 and vulkan llama.cpp, AMD Software: Adrenalin Edition™ Drivers 26.10.08. Scenario consumption is modeled at 8 hours/day effective utilization, equivalent to approximately 0.573M output tokens/day and 5.73M input tokens/day, or approximately 6.3M total tokens/day. Electricity assumes 150W sustained “nightmare case” draw at $0.15/kWh for 24 hours/day, or $16.20/month. The $0.15/kWh assumption is a rounded estimate broadly aligned with U.S. Energy Information Administration February 2026 average retail electricity pricing , including 14.36¢/kWh for U.S. total all-sector pricing and 17.65¢/kWh for U.S. total residential pricing Source: https://www.eia.gov/electricity/monthly/current_month/april2026.pdf. Three-year local total includes upfront hardware cost plus 36 months of electricity; cloud total assumes equivalent token consumption over 36 months, pricing for Claude Sonnet 4.5, source: https://platform.claude.com/docs/en/about-claude/pricing. Pricing information fetched as of May 2026. Hardware price is based on AMD Ryzen™ AI Halo Box 128GB, as of 5/10/2026. Actual results may vary with workload, context length, caching, batching, model choice, electricity rate, hardware configuration, utilization, and real-world agent behavior.
- GD-243: Trillions of Operations per Second (TOPS) for an AMD Ryzen processor is the maximum number of operations per second that can be executed in an optimal scenario and may not be typical. TOPS may vary based on several factors, including the specific system configuration, AI model, and software version. GD-243.
Footnotes
© 2026 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, AMD ROCm, Radeon, Ryzen, and combinations thereof are trademarks of Advanced Micro Devices, Inc. Linux® is the registered trademark of Linus Torvalds in the U.S. and other countries. Windows is a registered trademark of Microsoft Corporation in the US and/or other countries. Other product names used in this publication are for identification purposes only and may be trademarks of their respective owners. Certain AMD technologies may require third-party enablement or activation. Supported features may vary by operating system. Please confirm with the system manufacturer for specific features. No technology or product can be completely secure.
- SHO-50: Testing as of May 2026 by AMD. Performance based on multiple runs of Pytorch-based workloads using default templates in ComfyUI 0.8.36. Workloads tested: Stable Diffusion XL, Flux Schnell, Flux 2. Klein, Qwen Image, Qwen Image Edit, Z Image Turbo, Ace Step 1.5, Ace Step 1.5 XL, Hunyuan 3D 2.1, LTX 2.3, and Wan 2.2. System specifications: Pre-production AMD Ryzen™ AI Halo developer platform with AMD Software: Adrenalin Edition™ 26.10.08 drivers vs. Apple Mac Mini with Apple M4 Pro processor, 12-core CPU, and 16-core GPU, MacOS Tahoe 26.4, 64GB memory and 512GB storage set to High Performance Mode. Performance may vary. SHO-5
- SHO-57: Testing as of May 2026 by AMD. Performance based on average across 3 runs of GPT OSS 120B model, Qwen 3.5 122B Model, Qwen 3.6B Model and GLM 4.7 Flash 30B Model. Testing completed on pre-production AMD Ryzen AI Halo developer platform with AMD Ryzen AI Max+ 395 processor, 128GB LPDDR5x memory, and Linux OS. Performance compared against NVIDIA DGX Spark, with 128GB LPDDR5x using latest software stack available as of May 6th 2026.Tokens per second calculated based on 100 token context. System manufacturers may vary configurations, yielding different results. SHO-57
- SHO-61: Testing as of May 2026. Retail price for NVIDIA DGX Spark is $4699, Retail price for AMD Ryzen AI Halo is $3999. Performance based on average across 3 runs of GPT OSS 120B model, Qwen 3.5 122B Model, Qwen 3.6B Model and GLM 4.7 Flash 30B Model. Testing completed on pre-production AMD Ryzen AI Halo developer platform with AMD Ryzen AI Max+ 395 processor, 128GB LPDDR5x memory, and Linux OS. Performance compared against NVIDIA DGX Spark, with 128GB LPDDR5x using latest software stack available as of May 6th 2026.Tokens per second calculated based on 100 token context. System manufacturers may vary configurations, yielding different results. SHO-61
- SHO-49: Specimen use-case using simplified assumptions. Cloud comparison assumes Claude Sonnet 4.5 standard API pricing of $3/M input tokens and $15/M output tokens, with a 10:1 input-to-output token utilization mix. Local scenario assumes AMD Ryzen™ AI Max+ sustained throughput at 128K context of 36 output tokens/sec and 446 input tokens/sec, measured on pre-production AMD Ryzen™ AI Halo developer platform using LM Studio 0.4.12 and vulkan llama.cpp, AMD Software: Adrenalin Edition™ Drivers 26.10.08. Scenario consumption is modeled at 8 hours/day effective utilization, equivalent to approximately 0.573M output tokens/day and 5.73M input tokens/day, or approximately 6.3M total tokens/day. Electricity assumes 150W sustained “nightmare case” draw at $0.15/kWh for 24 hours/day, or $16.20/month. The $0.15/kWh assumption is a rounded estimate broadly aligned with U.S. Energy Information Administration February 2026 average retail electricity pricing , including 14.36¢/kWh for U.S. total all-sector pricing and 17.65¢/kWh for U.S. total residential pricing Source: https://www.eia.gov/electricity/monthly/current_month/april2026.pdf. Three-year local total includes upfront hardware cost plus 36 months of electricity; cloud total assumes equivalent token consumption over 36 months, pricing for Claude Sonnet 4.5, source: https://platform.claude.com/docs/en/about-claude/pricing. Pricing information fetched as of May 2026. Hardware price is based on AMD Ryzen™ AI Halo Box 128GB, as of 5/10/2026. Actual results may vary with workload, context length, caching, batching, model choice, electricity rate, hardware configuration, utilization, and real-world agent behavior.
- GD-243: Trillions of Operations per Second (TOPS) for an AMD Ryzen processor is the maximum number of operations per second that can be executed in an optimal scenario and may not be typical. TOPS may vary based on several factors, including the specific system configuration, AI model, and software version. GD-243.