AMD announced several new products and software updates across its client, gaming, and AI computing segments during CES 2026, including new processors for AI-capable PCs, a dedicated developer mini-PC, and an upcoming gaming chip. The company also confirmed its annual Advancing AI 2026 event, scheduled for July in San Francisco.
New Ryzen AI and Ryzen AI PRO Processors
AMD introduced the Ryzen AI 400 Series and Ryzen AI PRO 400 Series processors. These processors are designed for Copilot+ PCs and business laptops, respectively. They use the "Zen 5" CPU architecture, integrated AMD Radeon 800M Series graphics, and second-generation XDNA 2 NPUs.
AMD states the NPUs deliver up to 60 TOPS of AI compute.
The processors feature up to 12 CPU cores and 24 threads. AMD claims these processors offer a 1.3x increase in multitasking speed and are 1.7x faster in content creation compared to competitors.
Systems featuring the Ryzen AI 300 Series or the Ryzen AI 400 Series processors are scheduled for availability in the first quarter of 2026 from manufacturers including Acer, ASUS, Dell, HP, GIGABYTE, and Lenovo. Desktop PCs with the Ryzen AI 400 Series are expected in the second quarter of 2026.
Ryzen AI Max+ Series Expansion
AMD expanded its Ryzen AI Max+ Series with the announcement of the Ryzen AI Max+ 392 and Ryzen AI Max+ 388 processors. These processors integrate "Zen 5" cores, AMD Radeon 8060S Series graphics, and second-generation XDNA architecture NPUs. They are intended for thin-and-light laptops, workstations, and mini-PCs. Systems with these processors are anticipated to be available in the first quarter of 2026 from partners such as Acer and ASUS.
Ryzen AI Halo Developer Mini-PC
AMD unveiled the Ryzen AI Halo, a mini-PC designed for AI development. The system is an AMD-branded product intended for local AI model inference and development. It is positioned as a competitor to Nvidia's DGX Spark.
- Hardware and Specifications: The Ryzen AI Halo measures 5.9 x 5.9 x 1.7 inches (6 inches square and less than 2 inches tall). It is powered by a 120W Ryzen AI Max+ 395 APU (Strix Halo) with 16 Zen 5 CPU cores, 40 RDNA 3.5 GPU compute units (offering up to 56-60 TFLOPS at FP16), a 50 TOPS XDNA 2 NPU, 128 GB of LPDDR5x-8000 unified memory (256 GB/s bandwidth), and 2TB of storage. A higher-capacity 192 GB version with a Ryzen AI Max+ 495 APU is in preparation.
- AI Capabilities: AMD states the system can run local AI models with up to 200 billion parameters at 4-bit precision.
- Software and Connectivity: The system runs standard x86 operating systems (Windows and Linux) and is optimized for AMD ROCm software and AI developer workflows. AMD will provide preinstalled and online "playbooks" for common AI frameworks, including vLLM, Llama.cpp, Ollama, and ComfyUI. Connectivity includes Wi-Fi 7 and a 10 Gbps Ethernet NIC. USB-4 may support RDMA, though AMD has not confirmed this.
- Pricing and Availability: Pre-orders for the Ryzen AI Halo begin in June 2026, starting at $3,999. AMD states this is $700 less than Nvidia's DGX Spark.
In LLM inference (token generation), AMD claims the AI Halo is 4-14% faster than the DGX Spark. AMD also estimates that the Ryzen AI Halo can save developers $750 per month compared to cloud API usage, based on eight hours of daily local AI coding.
Ryzen 7 9850X3D Gaming Processor
AMD introduced the Ryzen 7 9850X3D desktop gaming processor. This processor is based on the "Zen 5" architecture and features second-generation AMD 3D V-Cache technology.
It includes 8 cores and 16 threads, operates with a boost frequency of up to 5.6 GHz, features 104MB of total cache, and has a TDP of 120W. The processor is available at a price of $499.
AMD stated the Ryzen 7 9850X3D offers a 400 MHz boost clock increase over its predecessor and provides an average 27% gaming performance improvement compared to the Intel Core Ultra 9 285K.
The company also noted that memory speed has minimal impact on performance for this chip, with less than a 1% FPS difference observed across DDR5-4800 and DDR5-6000 in a test of over 30 games. Systems with the Ryzen 7 9850X3D are expected to be available from OEMs and retail partners beginning in the first quarter of 2026.
Software and Ecosystem Updates
AMD announced several software updates:
- AMD ROCm 7.2: This open software platform now supports Ryzen AI 400 Series processors and is integrated into ComfyUI. Compatibility extends to both Windows and Linux, with PyTorch builds accessible via AMD software.
- AMD Software: Adrenalin Edition AI Bundle: A new feature designed to streamline local AI setup, providing tools for applications such as image generation, local large language models (LLMs), and PyTorch on Windows.
- AMD FSR "Redstone": This includes FSR Upscaling and FSR Frame Generation, which utilize machine learning to enhance visual quality and frame rates in games. These features are available in the AMD Software: Adrenalin Edition 25.12.1 driver. Additionally, FSR Radiance Caching, a developer preview aimed at improving ray tracing performance, is available on GPUOpen.com.
Ryzen AI MAX APU and Radeon AI PRO GPU Capabilities
AMD published a guide detailing how to enable the OpenClaw AI agent on its hardware, introducing two configurations: RadeonClaw (based on Radeon AI PRO GPUs) and RyzenClaw (based on Ryzen AI MAX SoCs).
Ryzen AI MAX+ APU: Supports 128 GB of memory, with up to 112 GB of VRAM allocatable to Radeon 8000S GPUs.
For single-agent operations with the Qwen 3.5 122B model, AMD reports these APUs achieve up to 19 Tokens/s. Multi-agent configurations support up to two agents with 95K context concurrency. For Qwen 3.5 35B workloads, AMD reports 45 Tokens/s performance and processing 10,000 input tokens in 19.5 seconds. The chips feature a maximum context window of 260K.
Radeon AI PRO R9700 GPU: This 32 GB RDNA 4 GPU can process 10,000 input tokens in 4.4 seconds and provides 120 Tokens/s performance. It has a maximum context window of 190K.
Up to four Radeon AI PRO R9700 GPUs can be combined in workstation setups, offering 128 GB of VRAM for running larger 128B models locally.
Advancing AI 2026 Event
AMD announced its annual Advancing AI 2026 event will take place from July 22-23 at the Moscone Center in San Francisco, California. The event is designed to bring together developers, researchers, partners, and customers to showcase advancements in AI infrastructure, architecture, and development. AMD CEO Dr. Lisa Su is expected to deliver the opening keynote.
Key areas of focus are projected to include:
- Detailed information on Instinct MI450 accelerators.
- Updates on Zen6-powered EPYC Venice CPUs.
- Information on next-generation Instinct MI500 accelerators, reportedly planned for 2027.
- Software updates related to AI development.
The event will include technical talks, networking opportunities, and hands-on sessions, allowing attendees to interact directly with AMD's AI researchers, engineers, and partners. This industry-focused event is scheduled after Computex 2026, where AMD's focus is anticipated to be on consumer products such as next-generation Ryzen desktop processors.