Lucebox
lucebox.com
Actively hiring
AI Compute & Inference
Edge AI
Servers & Data Center Hardware
As featured in
Follow
Overview
News
Salaries
Products
People
Growth
Offices
Financials
Funding and backers
Backers
1
Y Combinator
Momentum
15/100
in the news
Last sign of life
2 mo ago
a posting or a news item
Tractionpublic signals, read monthly
Web presence
top 7%
of 133.2M domains by links, Common Crawl
Hacker News stories
5
best: 20 points
Registers and programsofficial lists and records that name the company
| Legal entity | AGMI AI Inc. | its site |
Backed byfrom the funds' portfolios and accelerator catalogs
Y Combinator Investors in similar startupsfunds that back the companies most like this one
Similar startupsby what they do, their industry and backers
Wally by RunAnywhere brings frontier-scale open models to hosted infrastructure, alongside local GPU and NPU acceleration and open-source…
Run open models privately on your own computers. Actual connects your hardware for local AI inference.
An on-device LLM understands your entire life, then proactively offers to get your work done through computer use. Private by design. Free…
One platform to operate the full AI lifecycle. Training, inference, agents.
Featherless is a serverless AI inference platform that makes tens of thousands of open-source AI models instantly available through a…
Sail is the AI inference provider designed for agents at scale. Our API lets you deploy billions of tokens towards your hardest problems.
News
Lucebox Geometric: DeepSeek V4 Flash 0731 reaches 32.7 tok/s on AMD Strix Halo
DeepSeek V4 Flash 0731 on 128 GB AMD Strix Halo: 82/92 quality, up to 32.7 tok/s decode, and 173 tok/s sparse prefill at 60K context.
Read more
Report
Lucebox (AMD Radeon AI PRO R9700 + Strix Halo) Beats NVIDIA DGX Spark by 3.63x on DeepSeek V4 Flash Decode Speed
51.1 tok/s on a Lucebox with AMD Radeon AI PRO R9700 + Strix Halo versus our 14.09 tok/s average on one NVIDIA DGX Spark for the full 284B DeepSeek V4 Flash model.
Read more
Report
DeepSeek V4 Flash: 284B model, up to 32 tok/s on AMD Ryzen AI MAX+ 395
AMD-Powered Lucebox runs the 284B DeepSeek V4 Flash model locally on AMD Ryzen AI MAX+ 395 with 128 GB unified memory, reaching up to 32 tok/s decode and roughly 250 tok/s sparse prefill.
Read more
Report
Report
lucebox/optimizations/kvflash at main Luce-Org/lucebox
LLM speculative inference server for consumer & heterogeneous hardware - lucebox/optimizations/kvflash at main Luce-Org/lucebox
Read more
Report
Luce Spark: a 35B MoE on a 16 GB GPU, without the offload tax
A 33-35B MoE only fires 8 of 256 experts per token, but holding it on the GPU costs you all of them. Luce Spark pins the experts your traffic uses, offloads the rest to CPU, and decodes the whole token in one fused graph, so a 33-35B MoE fits a 16 GB card
Read more
Report
GitHub - Luce-Org/lucebox: LLM speculative inference server for consumer & heterogeneous hardware
LLM speculative inference server for consumer & heterogeneous hardware - Luce-Org/lucebox
Read more
Report
Offices
Where the company hires and what each office is for
Cities
1
Hiring now
0
Countries
1
Milan
Italy
HQ