{
  "id": 15557450,
  "title": "FuTransformer: FP8 prefill and INT4 decode for Llama-style inference on Hopper",
  "url": "https://urgent.news/2026/10/16/futransformer-fp8-prefill-and-int4-decode-for-llama-style-inference",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-16T17:22:01.000Z",
  "source": {
    "name": "arXiv cs.AI",
    "slug": "arxiv-cs-ai",
    "url": "https://arxiv.org/abs/2610.23796v1"
  },
  "original_language": "en",
  "account": null,
  "summary": "Transformer prefill reuses weights across tokens and is dominated by matrix multiplication and activation data movement; single-token decode has limited reuse and is dominated by weight and KV-cache reads. We present FUTRANSFORMER, a CUDA implementation of a Llama-style Transformer block for NVIDIA Hopper GPUs. It applies CODA's GEMM-plus-epilogue approach to inference, adding direct KV-cache…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}