Story

On air. The free wire runs 15 minutes behind live. See pricing

World, crypto, finance and tech news, scored for impact.

CryptoTokenPostKorean, translatedFiled 19 h ago

13 models run on AMD GPUs… Preview backend requires NPU

Original headline13개 모델 AMD GPU서 실행…NPU 필요한 프리뷰 백엔드

Open the sourceReported by TokenPost, tokenpost.kr, 20 November 2026, by 강이안

Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.

FastFlowLM released v1.0.13, adding an oga-hipep backend that lets 13 AI models run on AMD GPUs on Windows via ONNX Runtime GenAI and AMD's hip-ep runtime, shipped in the standard Windows ZIP and MSI packages. GPU-tagged models carry the -oga-hipep suffix and can be launched with the flm run or flm serve commands, with the roughly 137MiB hip-ep runtime downloaded to the user profile on first use. The catch is that GPU execution still requires an NPU, so the backend expands hardware support without removing the NPU dependency. It matters because it broadens where FastFlowLM's local inference stack can run on AMD hardware. What to watch next is whether the preview backend matures and whether the NPU requirement is lifted for GPU-only systems.

Muhabbit's AI wrote this summary in its own words from TokenPost's report and translated it from Korean. It can get details wrong, so check the source.

Named in this story

FastFlowLM, AMD (AMD), oga-hipep, ONNX Runtime GenAI, hip-ep

A linked name opens that asset's sentiment page.