RSS Anyway logo
HotLatestFollowingStatusAbout

lucebox.com

Sign in to follow
lucebox.com
RSSAtomJSON
  1. 01.
    Laguna XS 2.1 33B on a RTX 3090: 296 tok/s peak, 152 tok/s at 256K context
    lucebox.com · /sitemap.xml#/blog
    ▲ 19 · Jul 11 · 1 comment on HN
  2. 02.
    A 35B MoE on a 16 GB GPU, without the offload tax
    lucebox.com · /sitemap.xml#/blog
    ▲ 14 · Jun 8 · 0 comments on HN
  3. 03.
    Gemma 4 26B edges out DeepSeek V4 Flash (284B) on ds4-eval-92, at 5x the speed
    lucebox.com · /sitemap.xml#/blog
    ▲ 0 · May 20
  4. 04.
    Launch and tune Lucebox with real agent harnesses
    lucebox.com · /sitemap.xml#/blog
    ▲ 0 · May 15
  5. 05.
    DFlash + PFlash on AMD Strix Halo: 2.5× end-to-end vs llama.cpp HIP
    lucebox.com · /sitemap.xml#/blog
    ▲ 0 · May 12
  6. 06.
    Laguna XS.2 on a 3090: 111 tok/s, 5.4x prefill, first MoE target for PFlash
    lucebox.com · /sitemap.xml#/blog
    ▲ 0 · May 8
  7. 07.
    PFlash: 10× prefill speedup over llama.cpp at 128K on a RTX 3090
    lucebox.com · /sitemap.xml#/blog
    ▲ 0 · Apr 28
  8. 08.
    DFlash on ggml: up to 207 tok/s Qwen3.5-27B on a RTX 3090
    lucebox.com · /sitemap.xml#/blog
    ▲ 0 · Apr 16
  9. 09.
    NVIDIA GPUs Work on macOS Again. We Benchmarked Them. The Driver Is a Miracle. The Inference Is Not.
    lucebox.com · /sitemap.xml#/blog
    ▲ 0 · Apr 14
  10. 10.
    Megakernel: Matching Apple Silicon Efficiency at 2x the Throughput on a RTX 3090
    lucebox.com · /sitemap.xml#/blog
    ▲ 0 · Apr 13