Lucebox AI
Lucebox aims to make powerful AI something people own rather than rent by co-designing local-AI hardware and software. It builds private, always-on inference workstations and an open-source engine for running large open models locally.
At a Glance
- Individual developers and technical power users
- Small development teams
- Organizations with privacy-sensitive code, files and research
- AI/ML engineers and open-model researchers
- +1 more
AI Tools by Lucebox AI
(1)
Lucebox
Local LLM Inference Server
Discussions
No discussions yet
Be the first to start a discussion about Lucebox AI
Latest News
PFlash v2: 3x faster processing on long prompts
Lucebox cluster: DeepSeek V4.1 Flash at up to 103 tok/s on two boxes and one 100 GbE link
Running DeepSeek V4.1 Flash on the Lucebox memory hierarchy: up to 3x faster with the R9700
Vision LLM inference: Lucebox reports 3.2x the throughput of NVIDIA DGX Spark
Products & Services
An 11.97-liter local-inference workstation with AMD Ryzen AI MAX+ PRO 495, 128 or 192 GB unified memory, AMD Radeon AI PRO R9700 with 32 GB GDDR6, 2 or 4 TB NVMe, Ubuntu Server, preinstalled Lucebox Engine and tuned models.
Apache-2.0 inference engine and local HTTP server for heterogeneous hardware and consumer GPUs, with OpenAI- and Anthropic-compatible APIs.
The workstation's local browser management interface and companion connector for enabling supported laptop applications and pointing them at the private local endpoint.
Market Position
Lucebox positions itself between cloud AI services and general-purpose local machines: it sells a pre-tuned, private inference workstation with an open-source engine, rather than metered hosted inference or a DIY GPU setup. Its own comparisons reference NVIDIA DGX Spark and Apple Mac M5 Max, while the engine competes technically with llama.cpp-based serving stacks by emphasizing heterogeneous memory, speculative decoding, custom kernels and turnkey deployment.
Leadership
Founders
Alessandro Puppo
Co-founder and CEO; YC P25 alumnus and formerly co-founder of Cua (YC P25), with prior experience at Notion.
Davide Cifarelli
Co-founder and CTO; previously founded Klara Labs, an open-source LLM-interpretability project.
Executive Team
Alessandro Puppo
Co-founder & CEO
YC P25 alumnus; formerly co-founder of Cua and previously worked at Notion.
Davide Cifarelli
Co-founder & CTO
Previously founded Klara Labs, an open-source LLM-interpretability project.
Founding Story
Lucebox was started around the view that AI should be owned and controlled locally instead of rented from cloud providers. Its initial vision combines purpose-built hardware, an inference engine tuned to that hardware, and open models so private AI can be fast and ready out of the box.
Business Model
Revenue Model
Up-front hardware sales for Lucebox workstations; the included local engine and models run without per-token usage charges. The company also describes a future option to buy a device or pay monthly for personal intelligence.
Pricing Tiers
Limited-time price; list price $6,999.
Available configuration upgrade.
Available configuration upgrade.
Mellanox ConnectX-5 CX556M kit, installed and configured.
Target Markets
- Individual developers and technical power users
- Small development teams
- Organizations with privacy-sensitive code, files and research
- AI/ML engineers and open-model researchers
- Teams running coding agents, research agents and local inference services
- Private coding agents and software development
- Always-on local research bots, automations and agent teams
- Running large open models and long-context workloads locally
- Multi-user text and vision inference
- Developers and organizations that want predictable hardware costs without per-token cloud bills