AI at AMD

Code local. Build faster. Have fun. 😎 Get a preview of our Vibe Coding with Local Models workshop, where AI agents work together to help you build something you'll actually want to play. See you at #AdvancingAI. #AMDevs

The Developer Zone is where ideas become code. Here, #AMDevs can get hands-on with interactive demos, AI labs, and the latest developer tools running on AMD hardware. Don't forget to complete labs, earn points, and win prizes. Explore it all at #AdvancingAI: amd.com/en/corporate/e…

Introducing Kimi K3: Open Frontier Intelligence 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost 🔹 Built for long-horizon agentic coding and self-evolving workflows Kimi K3 is now live on on Kimi.com, Kimi Work, Kimi Code, and the Kimi API. Open Weights by July 27, 2026. 🔗 API: platform.kimi.ai 🔗 Tech blog: kimi.com/blog/kimi-k3

ROCm 7.14 is here. Meet TheRock, the new production-ready open-source build and release system for ROCm, alongside expanded AI hardware support, updated frameworks, stronger developer tools, and simplified deployment. See what's new and start building: rocm.blogs.amd.com/ecosystems-and…

@AnushElangovan @radixark @sgl_project @digitalocean 👏

Great partnership with @radixark / @sgl_project and @digitalocean on MI350.

.@radixark brought @deepseek_ai V4 to production with DigitalOcean and @AMD. AMD Instinct™ MI350X GPU Droplets, plus joint engineering that drove a ~10x throughput gain. When the team that builds the inference engine picks your stack, that's the signal.

Registration Is Open for the AMD AI DevMaster Hackathon! Join the global online competition for developers, researchers, students, AI practitioners, and open-source contributors. Build innovative AI applications accelerated by AMD Radeon™ GPUs and the ROCm™ software stack and compete for a share of the $30,000 USD prize pool. Why Join? $30,000 USD in prizes Free AMD Radeon™ GPU access for eligible participants Three innovation tracks: Agentic AI, Multimodal AI, and Physical AI Compete solo or in teams of up to three Bring your ideas to life, showcase what you can build with AMD and compete with innovators from around the world. Register today and start building. luma.com/amd-4dhi

Better HIP kernels through synthetic data, multi-agent search, and reinforcement learning. See how researchers at @Stanford's Scaling Intelligence Lab are advancing code generation for AMD GPUs. scalingintelligence.stanford.edu/blogs/hipkerne…

Build practical ROCm skills through hands-on training in AMD GPU architecture, programming, and performance optimization. At #AdvancingAI, Alex Kuznetsov, PhD will lead the ROCm Certification Associate workshop, covering ROCm fundamentals, PyTorch, HIP programming, CUDA porting, profiling, and optimization. Participants will complete guided labs and a certification exam designed to help them build and optimize GPU-accelerated applications on AMD platforms. Register to attend: amd.com/en/corporate/e… #AMDevs

AMD is helping advance academic AI and HPC research through cutting-edge compute, scalable infrastructure, and expert collaboration. At #AdvancingAI, Tom Papatheodore will lead a workshop highlighting how the AMD University Program supports next-generation academic innovation, including real-world university research on AMD platforms and resources like the AI & HPC Cluster program. Register to learn how AMD is accelerating discovery across academic AI and HPC research: amd.com/en/corporate/e… #AMDevs

Join us for a deep dive into ATOM, AMD's high-performance plugin that brings native Instinct GPU optimizations directly into the vLLM serving stack — no forking, no rewrites. ATOM delivers AMD-optimized attention backends (via AITER), custom kernel tuning, and support for the latest MoE and multimodal models (including Qwen3.5 and Kimi-K2.5), while preserving full vLLM API compatibility. Join on Discord: discord.com/invite/amd-dev

🚀 New blog: Bringing DeepSeek-V4 Flash RL Training to AMD Instinct MI355X GPUs with Miles DeepSeek-V4 Flash RL now runs end-to-end in Miles on @AMD Instinct MI355X GPUs with ROCm! Together with AMD, we aligned the model behavior across SGLang rollout and Megatron training, validated over 100+ optimizer steps on four MI355X nodes: 1️⃣ Train-rollout log-prob gap bounded at ~0.09 across 100+ steps 2️⃣ AIME-2024 pass@1 rose from 0.39 to 0.49, pass@8 from 0.53 to 0.67 3️⃣ FP8 rollout + BF16 actor with datatype-aware online weight updates 4️⃣ Stable TP1 / PP4 / EP4 layout on ROCm without collective stalls 5️⃣ Hybrid attention, mHC mixing & hash-routed MoE aligned across both engines

This is the OpenClaw workshop you've been waiting for. 🦞 Build your own AI agent from scratch in under 45 minutes using open-source multimodal models. No experience required, just bring your curiosity. Seats are limited, so register now: amd.com/en/corporate/e… #AdvancingAI #AMDevs

Universities are rapidly expanding hands-on AI education, but many still face challenges around infrastructure, expertise, and access to scalable compute. At #AdvancingAI, Jens Stapelfeldt and Naveen Purushotham will share how the AMD University Program helps educators bring real-world AI workflows into the classroom through ready-to-deploy curricula, practical tutorials, and powerful compute platforms. Register for the session to learn how AMD is supporting the next generation of AI education and research: amd.com/en/corporate/e… #AMDevs

Come for the workshops. Stay for the leaderboard. 🏆 Complete developer activities, rack up points, and compete for exclusive AMD swag and prizes, including an AMD Ryzen AI Halo developer platform, AMD Ryzen AI 7 Pro 350 laptop, and AMD Radeon RX 9070 XT GPU. Let the quest begin. Register today: amd.com/en/corporate/e…

Spend time building, not configuring. AMD Ryzen™ AI Halo gives developers the power to run larger models and more complex workflows locally without being locked into a single deployment approach. Get started: amd.com/en/products/pr…

Great spending time with Rick Mershad, CEO of @microcenter, to talk about an exciting milestone for builders. Ryzen AI Halo systems are now shipping from Micro Center. 💪 @AMD is the #1 desktop CPU at Micro Center with Ryzen X3D. 🎮 Radeon powers approximately 1 in every 3 desktop gaming GPUs sold at Micro Center. Now we are extending that same momentum into a new era. Ryzen AI Halo gives AI developers and builders the ability to run larger models locally, iterate faster, and unlock powerful new AI experiences while keeping more of their work on device for greater privacy and security, without cloud limits. The Agentic PC is here 🚀

What if you could learn an entire AI inference workflow—from training a model to deploying it and benchmarking its performance—in a single hands-on tutorial? In this tutorial, you'll take a CTR prediction model through every step of the journey using AMD Instinct™ GPUs and the ROCm software stack. You'll learn how to: How to obtain a ROCm-enabled Triton Inference Server container image. How to deploy an ONNX model on Triton Inference Server using the ONNX Runtime backend accelerated by the MIGraphX execution provider. How to measure the server’s throughput and latency on your own hardware with perf_analyzer, and how to interpret the results. This beginner tutorial uses a real click-through rate (CTR) prediction model called FinalNet and walks you through the process end-to-end: the model is trained first, then deployed and benchmarked. Check out the tutorial: rocm.docs.amd.com/projects/ai-de…

Whether you're exploring scaling LLM deployments, experimenting with agentic workflows, or enterprise AI there's a workshop for you at #AdvancingAI. The Developer Track features hands-on sessions designed to help developers build practical skills, learn from AMD experts, and explore real-world AI workflows. Check out this week's featured workshops and start building your schedule today. Register now: amd.com/en/corporate/e… #AMDevs

Follow your passion. Follow your instinct. Follow what moves you. @gaeldelalleau, Co-Founder and CEO of @Kog__AI, shares advice for developers entering the AI era.