AMD

Ryzen AI Software 1.8.0 expands LLM and VLM support with eight new additions including Hunyuan-0.5B/1.8B/4B/7B-Instruct, HY-MT1.5-1.8B/7B translation models, VideoChat-Flash-Qwen2.5-7B and InternVideo2-1B video models, and Qwen3-VL-4B-Instruct. The release also adds TTS support (neutts-air, neutts-nano), embedding mode Stable Diffusion capabilities grew with new models and dynamic resolution support, including SD1.5 i2i-canny, SDXL-turbo/base, SSD-1B, Playground v2.5, DreamShaper-XL-Lightning, FLUX.1-Schnell, SD3.0/3.5-medium, SD3.0-inpainting, and Flux.2-Klein-4B, with performance optimizations via -O1 for SD3.x and support for runn

AMD Enterprise AI version 2.2.2, released July 21, 2026, is a patch on top of 2.2 containing installation-package fixes only with no new features. The release is validated on ROCm 7.2.3 for Instinct and ROCm 7.13 for Radeon, and ships unchanged component versions: AI Workbench v2.0.1, AIM Engine v0.2.5, and Resource Ma Technical changes include a fixed first-boot race condition where the GPU operator config could apply before its CRD registered, now self-healed by ArgoCD retries, and added support for a custom Envoy Gateway HTTPS port. Known issues document deprecated v1alpha1 AIM Engine templates (migration to v1alpha2 profiles requ